Abstract. Fine particulate matter (PM_2.5) was collected in situ from peat smoke during the 2015 El Niño peat fire episode in Central Kalimantan, Indonesia. Twenty-one PM samples were collected from 18 peat fire plumes that were primarily smoldering with modified combustion efficiency (MCE) values of 0.725–0.833. PM emissions were determined and chemically characterized for elemental carbon (EC), organic carbon (OC), water-soluble OC, water-soluble ions, metals, and organic species. Fuel-based PM_2.5 mass emission factors (EFs) ranged from 6.0 to 29.6 g kg⁻¹ with an average of 17.3 ± 6.0 g kg⁻¹. EC was detected only in 15 plumes and comprised  ∼ 1 % of PM mass. Together, OC (72 %), EC (1 %), water-soluble ions (1 %), and metal oxides (0.1 %) comprised 74 ± 11 % of gravimetrically measured PM mass. Assuming that the remaining mass is due to elements that form organic matter (OM; i.e., elements O, H, N) an OM-to-OC conversion factor of 1.26 was estimated by linear regression. Overall, chemical speciation revealed the following characteristics of peat-burning emissions: high OC mass fractions (72 %), primarily water-insoluble OC (84 ± 11 %C), low EC mass fractions (1 %), vanillic to syringic acid ratios of 1.9, and relatively high n-alkane contributions to OC (6.2 %C) with a carbon preference index of 1.2–1.6. Comparison to laboratory studies of peat combustion revealed similarities in the relative composition of PM but greater differences in the absolute EF values. The EFs developed herein, combined with estimates of the mass of peat burned, are used to estimate that 3.2–11 Tg of PM_2.5 was emitted to atmosphere during the 2015 El Niño peatland fire event in Indonesia. Combined with gas-phase measurements of CO₂, CO, CH₄, and volatile organic carbon from Stockwell et al. (2016), it is determined that OC and EC accounted for 2.1 and 0.04 % of total carbon emissions, respectively. These in situ EFs can be used to improve the accuracy of the representation of Indonesian peat burning in emission inventories and receptor-based models.