Transport of nitrogen (N) and phosphorus (P) in runoff was measured using a rainfall simulator on a cotton hill-furrow system with a range of on-ground cover (0-60%), each with and without prior wheel traffic in the furrow. Total N and P losses in runoff (kg/ha) from the bare plots on a whole-field basis (wheel and non-wheel tracks) were equivalent to 1.5% and 1%, respectively, of fertiliser applied that season (218 kg N/ha, 27 kg P/ha). This was for a single runoff event of 17 mm, about 1/10 of the total runoff expected in a season. Retaining surface cover and avoiding wheel traffic were both effective in reducing runoff losses of total N and P, especially when used together. Retaining cover gave considerably lower concentrations of total P, and of N and P associated with sediment, with no significant differences (P > 0.05) between wheel and non-wheel tracks. The majority of nutrients were transported with sediment, for P for all treatments, and for N from low cover plots. Concentrations of dissolved N, dominantly as NO3-N, were unaffected by cover on non-wheel tracks but increased with cover on wheel track plots where runoff occurred as shallow interflow. On a whole-field basis, N was mainly in dissolved form at higher covers, because most runoff came from wheel tracks where interflow occurred. Reducing the ratio of wheel tracks to non-wheel tracks will reduce runoff of N and P. Interflow or exfiltration above an infiltration throttle layer is a worst-case scenario for runoff transport of soluble, poorly sorbed chemicals such as NO3-N, which would otherwise leach and not enter runoff. To improve water quality, for both sorbed and dissolved forms, the combination of retaining cover and avoiding wheel traffic and subsoil compaction is needed. Similarly, land uses involving high nutrient inputs should be avoided on soils with shallow subsurface restrictions to infiltration, which are thus prone to interflow. Primary (dispersed) clay and silt were slightly enriched in sediment in runoff, while primary fine and coarse sand were depleted. However, sediments were not enriched in total N and P compared with the soil surface, and organic carbon was only slightly enriched (enrichment ratio 1.06). This is typical of the behaviour of well-aggregated soils of high clay content.