Haze episodes occurred in Beijing repeatedly in 2013, resulting in 189 polluted days. These episodes differed in terms of sources, formation processes, and chemical composition and thus required different control policies. Therefore, an overview of the similarities and differences among these episodes is needed. For this purpose, we conducted one-year online observations and developed a program that can simultaneously divide haze episodes and identify their shapes. A total of 73 episodes were identified, and their shapes were linked with synoptic conditions. Pure-haze events dominated in wintertime, whereas mixed haze-dust (PM2.5/PM10 < 60%) and mixed haze-fog (Aerosol Water/PM2.5 ∼ 0.3) events dominated in spring and summer-autumn, respectively. For all types, increase of ratio of PM2.5 in PM10 was typically achieved before PM2.5 reached ∼150 μg/m(3). In all PM2.5 species observed, organic matter (OM) was always the most abundant component (18-60%), but it was rarely the driving factor: its relative contribution usually decreased as the pollution level increased. The only OM-driven episode observed was associated with intensive biomass-burning activities. In comparison, haze evolution generally coincided with increasing sulfur and nitrogen oxidation ratios (SOR and NOR), indicating the enhanced production of secondary inorganic species. Applicability of these conclusions required further tests with simultaneously multisite observations.