The fate and transport of endosulfan (6,7,8,9,10,10-hexachloro-1,5, 5a,6,9,9a-hexahydro-6,9-methano-2,4,3-benzodioxathiepin 3-oxide) applied to cotton (Gossypium hirsutum L.) fields were studied throughout three consecutive years on two selected locations in New South Wales (Australia). Rates of dissipation from foliage and soil, volatilization from the field, and transport of residues in irrigation and/or storm runoff waters were measured in order to estimate a total field balance. Dissipation of endosulfan from both foliage and soil is best explained by a two-phase process rather than by a first-order decay. Half-lives of total endosulfan toxic residues (alpha- and beta-endosulfan and the sulfate product) in the first phase were 1.6 d in foliage and 7.1 d in soil, and could be explained by the rapid volatilization of the parent isomers in the first 5 d (up to 70% of endosulfan volatilizes). In the second phase, half-lives were 9.5 d in foliage and 82 d in soil, mostly due to the persistence of the sulfate product. Concentration of endosulfan residues in runoff water varied from 45 to 2.5 microg L(-1) depending on the residue levels present on field soil at the time of the irrigation or storm events. These in turn are related to the total amounts applied, the cotton canopy cover at application, and the time since last spraying. Most of the endosulfan in runoff was found in the water phase (80%), suggesting it was bound to colloidal matter. Total endosulfan residues in runoff for a whole season accounted for no more than 2% of the pesticide applied on-field.