The Myanmar
Conflict Monitor

1. Broad-Based Contraction in Event Counts Contrast with Intensifying Lethality and Fatality Concentration

Total conflict events declined by 35.4 percent—a drop observed across all event categories rather than armed clashes alone—yet total fatalities decreased by only 23.0 percent. Fatalities concentrated heavily in September, where a modest 11.9 percent increase in events produced 1,257 deaths (a 44.3 percent surge over August), making it the deadliest month of the quarter.

2. High-Fatality Engagements Shifted Toward Riverine Logistics Corridors and Rakhine State

Sagaing and Magway regions saw their combined share of national fatalities drop to 27.8 percent, as military supply flotillas along the Ayeyarwady River to Bhamo and offensives into ULA/AA-held Ponnagyun Township shifted the geographical center of gravity. Four of the quarter's ten deadliest incidents occurred in September along the Bhamo–Shwegu corridor and in Ponnagyun, with Shwegu Township recording 304 fatalities—the highest death toll of any township nationwide.

3. Record High Ratio of Airstrikes per Clash as Military Remains Primary Perpetrator of Civilian Deaths

While armed clashes fell by 34.7 percent, airstrikes within armed conflict reached 231 events, pushing the ratio of airstrikes per clash to 0.349—the highest level documented since Q1 2021. The Myanmar military was the perpetrator in 85.3 percent (622 deaths) of all 729 recorded civilian fatalities, highlighted by a September 28 airstrike on the central market in Kyauktaw that killed 50 civilians.

In July, military operations prioritized securing supply lines to isolated outposts. On July 8, a military flotilla of about 28 vessels transporting a year’s worth of rations to Homalin ascended the Chindwin River, shelling villages on both banks as it transited Mingin Township. Residents of over 30 villages fled, generating approximately 20,000 recorded displacements. A comparable pattern of securing riverine corridors with waterborne convoys re-emerged on the Ayeyarwady River in September.

On July 31, a military push to control the Kalay–Tamu highway in Kale Township triggered 30,000 displacements—the single largest displacement event of the quarter. Displaced populations faced urgent shortages of food, tarpaulins, and antimalarial medication amid the monsoon season.

August was characterized by lethal airstrikes on religious compounds where civilians had gathered, alongside military counter-offensives targeting economic assets. On August 10, a jet circled for approximately 20 minutes over Thandwe Township, Rakhine State, before dropping three bombs near a village monastery, killing 16 civilians, including children. On August 21, in Myaung Township, Sagaing Region, an airstrike hit a Buddhist meditation center where approximately 110 elderly villagers were observing the sabbath, followed minutes later by a second strike; 15 civilians were killed.

On August 14, an estimated 700-strong military column with close air support recaptured two oilfield tracts in Myaing Township, Magway Region, that had been out of military control for over two years, leaving roughly 8,000 displacements ongoing near the fields into September. Concurrently, in Mohnyin Township, Kachin State, the Shanni Nationalities Army (SNA) and allied forces carried out house-to-house and gold-panning site raids around Indawgyi Lake on August 21, forcibly recruiting at least 600 individuals at a quota of one person per household.

In September, high-intensity combat shifted to the Ayeyarwady river corridor and Rakhine State. On September 2, a convoy of roughly 15 vessels and ground escort columns heading to Bhamo was ambushed in Shwegu Township by the KIO/KIA, PDF, and allied forces. The largest vessel was sunk, and 151 people were reported captured, though military sources denied any captures had occurred. On September 13 in Bhamo Township, KIO/KIA-led forces routed a military column of about 100 composed primarily of conscripts mobilized under the February 2024 Conscription Law, capturing over 30.

In Rakhine State, the military launched a coordinated offensive on September 5, deploying approximately 1,000 troops across five columns supported by naval artillery, airstrikes, and drones into ULA/AA-controlled Ponnagyun Township. Later, on September 28, military jets bombed the central market—comprising over 200 shops—and an adjacent bridge in Kyauktaw town, causing mass civilian casualties that included entire families and local Rohingya residents.

Total conflict events fell by 35.4 percent compared to Q2 2026. This contraction occurred across all major event categories rather than armed clashes alone. Nevertheless, three primary categories—violence against civilians, armed conflict, and coercion—continued to account for over three-quarters of all recorded events nationwide (Table 1). Monthly event totals held steady at suppressed levels throughout the quarter, demonstrating that the lower quarterly total reflected a sustained trend rather than an isolated single-month drop (Figure 1). Unidentified armed groups were the acting party in 4.4 percent of events (compared to 4.5 percent in Q2).

Direct clashes between combatants dropped by 34.7 percent to 661 events, while mine and IED incidents fell to 64 events. In contrast, airstrikes within armed conflict rose to 231 events, indicating a continued reliance on standoff airpower even as ground-level engagements receded.

"Fire on civilian targets" totaled 779 events, surpassing armed clashes (661 events) to become the single most frequent sub-category of the quarter. These incidents rose to 347 events in September alone. This late-quarter surge in standoff strikes on civilian areas underscored an intensification of operational activity heading into the final month.

Forced recruitment and portering accounted for 296 events, spiking to 131 events in September. This late-quarter rise coincided temporally with active military thrusts, including the Ponnagyun offensive and the frontline deployment of newly conscripted personnel observed in Bhamo.

Troop deployment and basing events rose steadily each month, reaching 53 events in September. Conversely, peaceful protests declined consecutively across the quarter, falling to 13 events in September.

A total of 3,004 fatalities were recorded in Q3 2026, a 23.0 percent reduction from Q2 that brought the daily average down from 42.9 to 32.7 deaths (Figure 2).

However, lethality remained concentrated: the 100 deadliest events (roughly 2.6 percent of all events) accounted for 60.7 percent of total fatalities (Figure 3). Furthermore, four of the quarter's five deadliest single days occurred in September, reflecting a surge in operational intensity toward the end of the period (Table 3).

The Myanmar military and affiliated forces suffered 2,033 fatalities, representing 67.7 percent of the quarterly total. Military-side deaths rose sharply in September, reaching 930 deaths compared to 551 in August (Table 2).

This escalation was driven by heavy engagements along the Bhamo–Shwegu river corridor and the Ponnagyun offensive. (Note: Some high fatality figures in major clashes rest on claims by the parties themselves; for example, 100 military deaths in Hpapun were reported by a resistance commander, while 100 deaths in Ponnagyun were reported by military sources).

Civilian fatalities fell by 16.8 percent to 729 deaths, declining less than total fatalities. In absolute terms, civilian deaths rose each month, climbing from 211 in July and 240 in August to 278 in September. The decline in the civilian percentage share of total deaths in September was caused by the surge in military combatant casualties rather than an actual drop in civilian deaths.

The military was the perpetrator in 622 civilian deaths (85.3 percent), with the vast majority occurring within "violence against civilians" events. In accordance with the project codebook, civilian fatalities include local ward and village tract administrators targeted as a result of their administrative positions.

Fatalities caused by mines and IEDs dropped by 63.7 percent to 148 deaths (down from 408 in Q2). This reduction outpaced the drop in recorded mine incidents, marking a sharp decline in mine-related lethality during the quarter.

Conflict was documented across 251 townships (−8.4 percent), with 119 townships recording incidents in all three months (Map 1).

Sagaing Region remained the most active area with 544 events, despite event counts falling by more than half (52.5 percent). Concurrently, Kachin State and Magway Region recorded the highest death tolls nationwide at 533 fatalities each, followed by Rakhine State with 398 fatalities.

Fatalities in Kachin State occurred predominantly among combatants during force-on-force engagements; of 533 total deaths in the state, 442 were military personnel.

These casualties were heavily concentrated along the Ayeyarwady River in Shwegu Township (304 deaths) and Bhamo Township (109 deaths) (Table 5). This corridor represents a critical logistical lifeline for the military to reinforce its garrison in Bhamo; on August 28, riverside villages in Shwegu Township, on this route, were hit by air and ground attacks.

In contrast to Kachin State, civilian deaths accounted for a high proportion of casualties in Sagaing Region: of 301 total fatalities, nearly half (148 deaths) were civilians.

This dynamic persisted because, despite fewer ground clashes, settlements remained subject to aerial bombings and artillery strikes, such as the August 21 monastery bombing in Myaung Township. In Yinmarbin Township, 23 events were documented without any recorded fatalities.

Rakhine State was the primary region experiencing an escalation in conflict events, rising by 25.1 percent. Sittwe (108 events) and Ponnagyun (91 events) ranked among the country's most active townships.

This escalation reflected a military transition from static defense to offensive maneuvers along the Sittwe–Ponnagyun corridor, placing civilians in ULA/AA-controlled northern Rakhine on the frontline. Rakhine State recorded 121 civilian fatalities, including those killed in the Kyauktaw market bombing. Reports also indicated that local Rohingya in Sittwe were detained, provided brief military training, and deployed at the front of advancing military columns (reported on September 11).

Kayah State and Yangon Region: Kayah State recorded a 35.2 percent rise in events, with 41 civilians among 99 fatalities. Yangon Region recorded a 17.6 percent increase in events, though only 5 fatalities were documented.

Ayeyarwady Region: Events fell by 33.0 percent. While 83 fatalities were recorded in Yegyi Township, most stemmed from the friendly-fire airstrike on a military column rather than escalating clashes.

Among territorial control change events, the military recaptured positions at an Indawgyi hilltop base (July 2–5), the Myaing oilfields (August 14), and a village along the Hpapun road (September 27). Conversely, the KNU/KNLA captured a military camp in Hpapun on September 8 that had been garrisoned by the military for over 30 years.

A total of 155 acting parties were documented during the quarter, with the Myanmar military recorded as an acting party in 2,488 events (Table 6). (Note: "Acting party" counts the events in which MCARI recorded an organization as the acting party and does not constitute a legal determination of responsibility).

Within armed conflict events, 82 parties were identified. Following the military and unidentified armed groups, the most active formations were the ULA/AA, KIO/KIA, Combined Group, and KNU/KNLA. (Note: Because People's Defense Forces are disaggregated by district and township names in the database, generic references to PDF do not capture the aggregate scale of local defense forces).

The quarter's deadliest battles coincided with the operational areas of three major ethnic resistance organizations: the ULA/AA in Ponnagyun and Ngape, the KIO/KIA along the Bhamo–Shwegu axis, and the KNU/KNLA in Hpapun.

Concurrently, the involvement of unidentified armed groups in armed conflict events fell to 18.2 percent (down from 23.0 percent in Q2), indicating that a larger share of armed clashes could be attributed to established organizations.

Disaggregating the data into two primary analytical layers highlights distinct forms of warfare operating in different parts of the country:

• Violence Against Civilians, Coercion, and Protests Layer: Accounted for 2,209 events and 663 civilian fatalities, concentrated predominantly in Sagaing and Mandalay regions.
• Armed Conflict Layer: Accounted for 1,136 events and 1,939 military combatant fatalities, concentrated primarily in Kachin State and Magway Region (Table 7).

This division demonstrates that while populations in Sagaing and Mandalay faced the brunt of standoff shelling, airstrikes, and detentions, combat casualties were concentrated in the armed conflict layer in Kachin and Magway.

Military-aligned armed groups participated in several notable operational events. The Shanni Nationalities Army (SNA) fought alongside the military in the Indawgyi area of Kachin State (July 2–5) and enforced a one-person-per-household forced recruitment drive on August 21. Additionally, Pyu Saw Htee militia units were recorded in 48 events during the quarter.

Of 34 targeted killing events, 12 were carried out by unidentified armed groups, resulting in 11 civilian deaths. Ward and village tract administrators were the most targeted category, accounting for 13 events and 10 civilian deaths (coded as civilians per codebook criteria).

Screening for changes in active armed groups showed:

• Newly Documented Formations: PDT (Madaya) with 6 events, DPLF/DPLA with 5 events (including the July 22 Ywangan ambush), and the Steering Council for the Emergence of the Federal Democratic Union with 5 events.
• Absences from Current Records: Formations active in Q2 that were not recorded as a main party in Q3 included the People’s Liberation Army (PLA), PDF (Pakokku District), and PDF (Yinmarbin District). Absence from reporting does not indicate that conflict ceased in these operational areas (e.g., Yinmarbin Township continued to record conflict events).

Nationally, 6,677 individuals were detained (−8.9 percent), declining consecutively across the quarter from 2,808 in July to 2,401 in August and 1,468 in September (Table 8). In accordance with the project codebook, detentions encompass not only political detentions, but also individuals forcibly conscripted, pressed into portering, or seized as human shields.

Consequently, single large-scale events accounted for the bulk of detentions in the two most affected townships (Figure 4):

• Mohnyin Township (892 individuals): Included the August 21 drive in the Indawgyi area in which at least 600 individuals were forcibly recruited under a quota of one person per household.
• Myawaddy Township (805 individuals): Included a July 25 raid on an online scam compound resulting in the detention of 777 individuals, predominantly foreign nationals.

These records demonstrate that aggregate detention figures reflect diverse forms of deprivation of liberty, combining political detentions with forced military recruitment and a raid on a scam compound.

A total of 291,643 displacements (−58.8 percent) were documented, with July accounting for 186,177 displacements. More than half of all displacements nationwide (52.8 percent) were concentrated in just eight events.

The largest among these occurred in Sagaing and Magway regions: 30,000 displacements from the highway offensive in Kale Township, 20,000 from Chindwin River shelling in Mingin Township, and two separate events of 20,000 displacements each in Chauk Township (resulting from arson and forced eviction). The primary triggers among events recording displacement were artillery and missile strikes (70 events), forced evictions (47 events), and property destruction (30 events).

Mass displacement generated acute humanitarian challenges: displaced families in Kale faced urgent shortages of food, tarpaulins, and malaria medication during the monsoon, while roughly 8,000 displacements persisted near the Myaing oilfields through September.

Defections and surrenders rose by 25.1 percent to 996 personnel—the only core indicator to increase—though monthly counts declined from 503 in July to 351 in August and 142 in September, with no single event accounting for over half of the total.

Conversely, releases dropped by 85.3 percent to 879 individuals, reflecting a sharp reduction in prisoner discharges compared to the preceding quarter.

Based on operational trajectories documented in Q3 2026, the following dynamics and uncertainties warrant continued observation:

1. Will the Reliance on Airstrikes Continue to Rise?

Airstrikes within armed conflict rose for four consecutive quarters starting in Q3 2025, pushing the airstrike-to-clash ratio to a record 0.349 as ground clashes fell. However, if clash counts rebound in future quarters, the ratio could decrease even if the volume of airstrikes remains high. Furthermore, this ratio does not include airstrikes carried out against civilians in non-combat settings.

2. Trajectory of High-Intensity September Battlefronts

Clashes along the Bhamo–Shwegu river corridor and operations in Ponnagyun drove a sharp rise in military combatant fatalities from 551 in August to 930 in September. However, some of these large death figures rest on claims by the parties themselves, and monthly fluctuations remain sensitive to outcomes in a small number of large engagements.

3. Monthly Increases in Civilian Fatalities

Despite an overall quarterly decline in civilian deaths, monthly fatalities climbed steadily from 211 in July and 240 in August to 278 in September. A three-month observation window is too brief to establish an enduring trend, requiring continued monitoring to evaluate whether civilian casualties continue to escalate.

4. Late-Quarter Escalation in Forced Recruitment

Forced recruitment and portering events surged from 82 in July and 83 in August to 131 in September, accompanied by the frontline deployment of newly conscripted personnel in Bhamo. However, this surge was observed only in the final month of the quarter, and the scale of conscription events varied widely, with a single event in Mohnyin accounting for roughly 600 individuals.