China Cannot Yet Hold a Beach on Taiwan

There was widespread concern in Washington that China’s extensive fleet of civilian ferries could concentrate a surprise landing of 60,000 to 400,000 soldiers on the coast of Taiwan. According to the IISS and Defense Priorities’ Lyle Goldstein, Taiwan’s entire active army, including conscripts, dropped from 240,000 in 2000, to 88,000 in 2020, and has only crept up to 95,000 in 2026.

We find first that the PLAN can land only a single Marine brigade and its supplies. While it can re-embark and land a second brigade within 72 hours, the amphibious flotilla will need to return to port and, in the meantime, strand the landed brigade, thereby being unavailable for emergency evacuations. Second, Taiwan can deploy from their garrisons onto a broad road-network with sufficient speed to counter-concentrate against landings on most of its Western coast, captured in this battalion-scale Vassal module. Third, China lacks an artillery substitute to interdict the movement of Taiwanese units and suppress fires because of Taiwan’s plentiful East-West roads.

Though the PLAN (People’s Liberation Army Navy) increased its Marine Brigades (BDEs) from six to eight in 2025 (IISS), their scarcity and specialized equipment suggest they will be reserved for the spearheading of amphibious attacks. Two of these BDEs are actually Army (PLA) units deployed with the 74^th Group Army in Southern Theatre Command. The PLA (Army) possesses two Air Assault Brigades and 14 Special Operations Brigades; most likely light infantry not trained for unconventional warfare. The Airborne Corps has five Airborne Brigades and a single Air Assault Brigade.  

The lethality of defensive fire on the landing beaches, the likelihood of landing craft being “holed” and clogging the constrained beach front, and the difficulty of unloading on an unprepared coast, leads us to conclude that while the fleet of civilian RO-RO ferries are capable of launching amphibious vehicles at sea, they will be withheld until the docking areas are cleared of enemy mortar fire and hill-top observed artillery bombardment. While these vessels may meet military standards and train with the PLAN, they are particularly vulnerable to damage and are more likely to wait to deploy with the aid of a secured port or mobile pier. This required safe distance can range from 3000 meters to 10 km. The PLA (Army) possesses an assortment of 258 landing craft, the great majority of which could not safely traverse the Taiwan Strait, and only five amphibious ships (Yunda LSTs and Yujiu LSMs).

Only the PLAN’s dedicated amphibious ships can force a landing, which is problematic given that we estimated its maximum lift capacity of one brigade. We found that the IISS lift estimates are in error because they group vehicles and personnel together, when their real capacity is either/or. Further, they don’t account for the wasted space between vehicles used for tie-down chains to prevent shifting with ship movement for access and for fire lanes. The U.S. Navy refers to this as “Broken Stowage.” We will be applying the values of the PLAN’s vehicle-heavy 3^rd Marine Brigade (based at Jinjiang), the best-equipped unit, in this simulation.

The 3Mar BDE is comprised of nine battalions (BNs), including four Combined Arms, detailed at links listed below. We calculated an additional 4038 STON of supplies for three days distributed on the lifted ground vehicles, eighty percent of which is fuel and artillery ammunition.

Lifted by: One Yushen Type 075 232m LHD, one Yuzhao Type 071 LPD, and their complement of 6 Yuyi LCACs and 32 Z-8 trooplift helicopters.

Lifted by: one Yuzhao Type 071 LHD with three Yuyi LCACs and 4 Z-8 helicopters.

Lifted by: 11 Yuting Type 072 LSTs and five Yunshu Type 073A LSMs.

Lifted by: seven Yuting Type 072 LSTs.

Lifted by: One Yuting Type 072 LST and five Yunshu Type 073A LSMs.

Lifted by: nine Yuting Type 072 LSTs

Lifted by: single Yuting Type 072 LST.

Unused are only three Yuzhao Type 071 LPDs, with a one-and-a-half battalion capacity, which will likely be carrying the 3Mar BDE supplies for the landing. It is assumed that the initial wave is carrying a combat load for just 24 hours of fighting, with two additional days of supplies aboard ship to be disembarked once the beach is secured. Following behind are three escorted 7-knot flotillas each of 125 civilian deck carrier LCT(S/L)s, in successive waves, one at the Strait centerline associated with the exercise deception, and the remainder proceeding from mainland Chinese ports. Each flotilla can carry one PLAGC (PLA Ground Force) Heavy Combined Arms Brigade (the 2^nd, 35^th, and 160^th BDEs of the Eastern Theatre’s 71^st Group Army) totalling over 18,000 personnel, which will be disembarked directly on the captured beach. These take 10 to 16 hours to cross the Straits, and will begin landing at H+5, H+8, H+10 in the closest Taoyuan scenario. As these are civilian transports lacking any defensive countermeasures, they are extremely vulnerable to artillery and tank fire, and will only land once a beach is cleared, which in the scenarios below, never happens.

The manner of a Chinese amphibious assault is to concentrate the launching platforms 21 nautical miles (nm) offshore of a targeted beach, just beyond the curvature of the earth, with ample air defense protection provided by naval ships and land-based aircraft. Assigning each ship a maneuvering and off-loading space of nearly one nautical mile (nm) squared results in a large flotilla footprint of 10 by 8 nm, not including air defense, surface defense, anti-submarine, anti-UUV, or minesweeping platforms. This is the minimum level of dispersal needed to manage ship, shore, air and submarine-launched anti-ship missile leakers, mines, UUVs, torpedoes, and collisions.

The manner of the amphibious attack is to make a coordinated launch with the slower assault vehicles first (15 knot ZBD-05, ZTD-05, and PPL-05, in a maximum sea state of 3), then the twenty-seven 80-knot Yuyi LCACs (hovercraft), and the 92 Z-8 ferrying trooplift helicopters (42 of these are borrowed from the PLAAF). The PLZ-07B artillery is amphibious but cannot be securely deployed until the beach is cleared of enemy fire and it is safe to establish a large artillery ammunition dump. In some scenarios detailed below, most of the LCACs and helicopters are destroyed. China will not deploy the Yuting LSTs until the beach has been cleared of Taiwanese mortars to a depth of 4 km. It takes approximately 6 minutes to flood the well decks, with vehicle loading and disgorging taking 30 minutes. It would take approximately two hours from the unloading into formation and travel to the beach over the 21nm. A PLANMC (PLAN Marine Corps) brigade beach frontage is 2-4 km, with 0.5-1 km per bn.

Once the beach is secure, the ten Yunshu LSMs and 29 Yuting LSTs, carrying the brigade artillery stockpile of shells and the support elements, will unload. Ideally these vessels would land at high tide and retract from the beach within two hours, but more likely will be re-floated by the high tide twelve hours later. We are assuming that without an extended campaign to secure air superiority consistent with this surprise scenario, a PLAAF cross-Straits heliborne or airborne landing near Taiwan’s main garrisons would have exceptionally high attrition rates. Permitting 200 meters between ships on a 4km beach, only 20 ships will be able to unload at the same time. That leaves 19 ships waiting for the next tide. However, several of the ships won’t be able to retract due to battle damage or broaching.

If the PLAN amphibious flotilla returns to port to load its second amphibious BDE, it will not be able to evacuate the 3^rd Marine BDE in the event of a defeat. Even presuming the simultaneous neutralization or at least suppression of Taiwan’s islands of Kinmen and Matsu opposite the Chinese inland ports of Xiamen and Fuzhou, the ease of Taiwanese interference with subsequent loading suggests that the flotilla will depart for safer ports as far away as Hong Kong and Shanghai. Heading toward either Wenzhou and Shanghai is likely too exposed to U.S. and Japanese interception. Shantou Naval Base is 250 nm away and Hong Kong is 450nm away, and loading a second brigade could take three days, which is how long the 3Mar BDE must hold the beach before reinforcement. It will take much longer if Taiwan interferes with the loading. The 3Mar BDE will be out of supply on the third day, assuming that they are relying on only PLAN amphibious shipping.

The PLANMC’s ability to secure its beach depends entirely on the speed of the ROCA’s (Republic of China Army) redeployment, especially given Taiwan’s plentiful artillery (2,104 tubes not including mortars), at the rate of one tube per 46 defending soldiers. This compares with a NATO Cold War ratio of one tube per 70-80 soldiers and a Soviet Motor Rifle Division ratio of one tube to 36 soldiers. While it is likely that the ROCA will disperse from their bases to likely local landing areas during the pre-war crisis period, we calculate their deployments from their permanent garrisons if decision-makers in Taipei choose a non-provocative stance, the most likely cause of surprise attacks according to Columbia University’s Richard K. Betts.

We assume that nighttime base redeployments will be slowed by 50 percent due to refugee traffic, with normal tracked/wheeled units covering 15/12 km/hr by day/night, reduced to 12km/10km day/night per hour for dispersion to avoid air attacks. We produced a ROCA arrival time-table for each of Taiwan’s army battalions and companies for 26 beaches, based on an 800-meter hexagonalized Vassal module, where we also depict the landings at four prospective sites. We count the time from the moment of PLAN begins disembarkation, and ROCA artillery and local reserve units are assumed to be on location if they are within 10 km of the beach. The main factor facilitating the ROCA’s quick response is Taiwan’s especially extensive road network with few chokepoints. We assume non-use of the highway system, in case of the destruction of the on-ramps. The chart also shows the cumulative Operational Lethality Index (OLI) of arriving ROCA units.

The map and chart reveal three important points. First, that the PLAN can indeed land and secure certain poorly defended stretches of Taiwan’s littoral, perhaps as a political demonstration. However, the ease of landing in these regions, mostly on the central Western coast, are inversely related to their usefulness, isolatability and access to port facilities. Second, the ROCA response is scalable against even a simultaneous Group Army (division equivalent) landing in a single sector, or multi-Group Army landing in multiple sectors, limiting the effectiveness of a PLAN deception to misdirect ROCA forces. This means we can accurately extrapolate from a single brigade landing. Third, most ports, especially the key harbors of Taizhong and Kaohsiung, are very heavily defended and physically defensible. For our purposes, we are focusing on Max Stewart’s scenario of a landing just to the West of Taoyuan International Airport, which is a sound beachhead for the objective of isolating Taipei from the rest of Taiwan.

Our third finding is that China lacks the necessary means to sustainably suppress the Taiwanese approaches to the landing area. While the PLA has less than a brigade’s worth of airlift capacity for the insertion of paratroops (based on advertised capacity and floorspace estimates) without prior suppression of enemy air defense, crossing the 150 km of open water of the Taiwan Strait would likely involve a very high rate of attrition. For the same reason, it is risky sending 200 of its Changhe Z-10 or 120 Harbin Z-19 attack helicopters for fear that light damage would impair their return trip. It is very unlikely the PLARF (PLA Rocket Force) would waste its 225 conventional SRBMs or its 126 CJ-10/100 cruise missiles on an area suppression mission, even if equipped with cluster munitions warheads. PLAN warships lack the gunnery of sufficient weight to have an area suppressive effect, and it is ill-advised to operate in the Taiwanese inshore because of anti-ship missiles.

China’s only artillery with the required range firing from the mainland is the PHL-16 MRL launching four 370mm 220km-range BRE6 rockets. Although satellite-guided with a CEP of 30m, this is not a useful capability when trying to suppress redeploying targets, and is at least ten times more expensive than its unguided version. Furthermore, whether and to what extent the PLA can secure drone surveillance over Taiwan is a function of their comparative technological innovativeness (as seen in the Russian-Ukrainian imbalance), which is not yet determinable. If Taiwan possessed Ukraine’s lead, it could even interfere with the PLAN’s re-loading of its amphibious ships in mainland ports. Even with drone reconnaissance, Russia has been unable to conduct deep interception of Ukrainian vehicles. Even with drone guidance, it is unclear how the PHL-16 would perform any better than an unguided suppression fire mission. We think it is far more likely that China will accept the losses associated with seeking suppression with low-flying JH-7A Feibao (NATO “Flounder”) fighter-bombers, each dropping twenty Type 250-3 bombs. In fact, the Flounder has not been seen carrying more than two six-bomb Multiple Ejector Racks. Equipping these aircraft with guided munitions will not improve the likelihood of identifying targets in Taiwan’s built-up and broken terrain. Nor has the PLAAF demonstrated a capacity for nighttime interdiction.

We apply the Admiralty Trilogy Group’s method of calculating target neutralization by bombardment. The unit of suppressible terrain used here is a 4-hectare target area (200-meters on a side), typically occupied by a force of company strength, fourteen of which fit into an 800-meter-wide hexagon. The total road-passable approach from the ROCA garrison to the beach landing at Taoyuan comprises an immense area of 384 square km, with no chokepoints, and comprising 8,400 targets. We shrink our impact area to the 4000-meter squared area of the landing zone comprising ten 800-meter hexagon areas, each of which will be required to suppress for the two-hour duration of the approach of the amphibious force and an additional hour for consolidation.

A PHL-16 firing four rockets, each with a 200kg warhead, creates only a 1.9 percent probability of suppression per target area. Assuming each PHL-16 can fire four rockets every ten minutes, for 140 minutes, the entire 140 PHL-16 force could produce a dismal 19% suppression in one 800-meter diameter hexagon for the period of bombardment. This capability can alternately be used to counter-battery Taiwanese artillery, but this will depend on sensors and drone reconnaissance. A single Flounder sortie producing a 5.4% suppression scaled-up to 200 airframes, produces a 54% suppression in one and half 800-meter diameter hexagons. For both cases, the low volume of fire creates an average target recovery time of thirty minutes, less than the time to generate a second sortie. These dismal results indicate that for an amphibious assault to suppress enemy targets requires the successful landing and supplying of the 3^rd Marine BDEs organic Artillery Bn.

We generate the ROCA combat power using the OLI formula, although WEI/WUV or any fair method of comparative quantification can be used as inputs for the ATC Model. We established the equipment for each of the four PLAN 3^rd Marine Brigade CA Maneuver Battalions and the Marine Recon Bn and calculated their OLI. For example, the final OLI calculation for an M1A1 MBT is Weapon OLI (478) * Mobility Factor (1.23) * Radius of Action (2.13) * Vehicle Punishment Factor (1.414) * Basic Load to ROF Ratio (0.71) * Fire Control (1) * Vehicle Movement Factor (0.95) = 1188 OLI.

We make a number of assumptions. First, that both PLAN and ROCA are rated as training “competent,” and ROCA reserve BNs are “inexperienced.” We decide that the intensity of the ROCA artillery interdiction of the beaches required that the PLAN advance 4 km inland to the Force Beachhead Line before it will land the LSTs to deploy the 3^rd Marine BDEs Artillery Bn: this was never achieved and the absence of this substantial source of counter-battery firepower is a key reason for the ultimate failure of the PLAN’s attack. Third, we believe that Marine-portable drones are far less effective in weight in producing damage per unit of time than the equivalent in mortar rounds, which is why we kept the mortars in lieu of drones for the PLAN. ATC calculates losses to OLI as well as personnel casualties. Dupuy developed a separate attrition model, also derived from the  TNDM, that we did not apply. Historically, personnel losses tend to be a fraction of equipment losses.

We ran three simulations of the 3^rd Marine BDE landing on the shores near Taoyuan Airport, the site used in Max Stewart’s scenario. In the first simulation, which is a baseline case, we assume the Taiwanese have been alerted at least two hours before Chinese debarkation and have deployed what we calculated is the minimum winning defending force that could successfully survive and hold the 3^rd Marine BDE on the beach. This combined-arms force comprises one reserve infantry, one Marine armor, and one artillery battalion, without any reinforcements. At 0600 the PLANMC hit the beach with only their recon and four maneuver BNs, fail in their attempts to push off the beach until 1000, and then collapse into a defense and initiate a failed evacuation that results in a surrender. Ten percent of landing craft are destroyed in the landing, and another 25 percent are lost during the beach withdrawal, typical for all of the cases below, highlighting the high attrition of an opposed landing against even a minimal defense.

In the second simulation, all ROCA units advance from garrison only at the moment of the PLANMC’s offshore debarkation, so only four reserve infantry and two artillery BNs are initially defending the beach. Within an hour, two artillery BNs and three brigades of mechanized infantry come in as reinforcements, followed by Marine and Army armor. This is its detailed explanation. In this more realistic and severe outcome, the PLANMC is abruptly blocked at the beach and counter-attacked and fails in an attempted evacuation to the sea, resulting in a PLANMC surrender.

In a third simulation, we posited landing three PLANMC Marine Brigades (3^rd, 4^th and 5^th BDEs), to examine whether China could overwhelm beach defenses with mass (despite China not having sufficient amphibious lift capability to drop a PLANMC division-equivalent 21nm off the Taiwanese coast). We assume that this scale of naval movement permits sufficient time for the ROCA to deploy to the coast (6 hours), reinforced by units as far away as 40 km, for a force total of four mechanized infantry, one Marine, two armored and two artillery brigades. The concentration of China’s scarce amphibious forces permitted the freeing up of Taiwanese reserves, countering the mass of the attack. For one hour the PLAN tried unsuccessfully to push off the beach, shifted to a defense, and initiated a failed evacuation in the fourth hour. While evacuations are more politically acceptable to Beijing than surrender, ad hoc efforts tend to escalate casualties.

In six more simulations, we modelled amphibious assault scenarios, depicted on the Taiwan Vassal map, against each of the coastal cities of Taipei (Brigade and Three-Brigade), Taizhong (Brigade and Three-Brigade) and Kaohsiung (Brigade and Three-Brigade). In these varied environments, including assaults on a coastal mountain (Taipei), port facilities (Taizhong), and urban terrain (Kaohsiung), the PLAN invariably suffers catastrophic defeats.

In all our simulations, at both the brigade and divisional-equivalent scale, and in varied operations against the south, center, north of Taiwan and against the capital, the PLAN suffered extreme losses similar to the Canadian experience at Dieppe on August 19, 1942.

We have four observations. First, that Taiwan is indeed defensible, even assuming the ROCA do not deploy from their garrisons until disembarkation. Nevertheless, ROCA units must practice rapid reaction under active interdiction as speed is the principal threat to any PLAN attack. Second, given that the initiative matters, Taiwan’s politicians must be willing to act pre-emptively despite domestic and diplomatic costs. Surprise outbreaks of war, such as the 1973 Arab-Israeli War, are often the result of politicians’ fear of overreacting or provoking the adversary, and not because the indicators of attack were undetected by intelligence.

Third, the evidence of the high cost of opposed landings suggests China is most likely to seize the sparsely populated region around the town of Chiayi and then conduct a months’ long buildup. Fourth, we expect that the PLA has recognized that it has a serious long-range fires problem and will seek a less expensive artillery solution which is not provided for by the PHL-16. While drones seem the immediate answer, their low rate of impact means that they would work best in a strategy of protracted conflict, and not one of seizing a beachhead.

William A. Sayers made his 30-year career in the Intelligence Community in Washington D.C. as a military analyst at DIA, CIA and the National Counterterrorism Center.  He holds Master’s degrees from the Marine Corps Command & Staff College, the Marine Corps School of Advanced Warfighting (where he was their only civilian graduate through at least 2011) and the Marine Corps War College.  He spent 2 years on active duty in the USAF and 3 years as a reserve intelligence officer in the Air Force Intelligence Service.  Sayers has published articles on the air war over North Vietnam in Air Power History and twice in Vietnam magazine.  He is currently the Principal Research Fellow and Director of Wargaming and Combat Simulations at The Dupuy Institute in Tyson’s Corner, Virginia.

Dr. Julian Spencer-Churchill is associate professor of international relations at Concordia University, and author of Militarization and War (2007) and of Strategic Nuclear Sharing (2014). He has published extensively on Pakistan security issues and arms control and completed research contracts at the Office of Treaty Verification at the Office of the Secretary of the Navy, and the then Ballistic Missile Defense Office (BMDO). He has also conducted fieldwork in Bangladesh, India, Indonesia, Egypt, and Taiwan, and is a consultant. He is a former Operations Officer, 3 Field Engineer Regiment, from the latter end of the Cold War to shortly after 9/11. He is currently collaborating with the Combat Modelling group at the Trevor Dupuy Institute.

The Central Intelligence Agency has reviewed this article and approved it for release to the public.  This does not constitute endorsement of its content.

image: PLAN Marine Corps command and staff personnel examine maps in the course of a cold weather training exercise in Inner Mongolia, March 2015. (Source: Xinhua)