Vegetative buffer strips show limited effectiveness for reducing antibiotic transport in surface runoff

Vegetative buffer strips (VBS) have been demonstrated to effectively reduce loads of sediment, nutrients, and herbicides in surface runoff, but their effectiveness for reducing veterinary antibiotic (VA) loads in runoff has not been well documented. The objective of this study was to determine the effectiveness of VBS vegetation and width on surface runoff loads of the VAs sulfamethazine (SMZ) and lincomycin (LIN). Experimental design of the plots (1.5 × 25 m) was a two‐way factorial with four vegetation treatments (tall fescue [Festuca aruninacea Schreb.]; tall fescue with switchgrass [Panicum virgatum L.] hedge; warm‐season native grass mix; and continuous fallow control), and four buffer widths (0, 2, 5, and 9 m). Turkey litter spiked with SMZ and LIN was applied to the source area (upper 7 m) of each plot, and runoff was collected at each width. Runoff was generated with a rotating boom simulator. Results showed VA loads in runoff at the 0‐m sampler ranged from 3.8 to 5.9% of applied, and overall VA transport in runoff was predominately in the dissolved phase (90% for SMZ and 99% for LIN). Among vegetation treatments, only tall fescue significantly reduced loads of SMZ and LIN compared with the control, with load reductions of ∼30% for both VAs. Estimated field‐scale reductions in VA loads showed that source‐to‐buffer area ratios (SBARs) of 10:1 to 20:1 reduced VA loads by only 7 to 16%. Overall, the grass VBS tested here were less effective at reducing SMZ and LIN loads in surface runoff than has been previously demonstrated for sediment, nutrients, and herbicides. Core Ideas Buffer width affected veterinary antibiotic (VA) load reduction more than vegetation treatment. Relative VA loads in runoff were similar to pesticide losses (3.8–5.9% of applied). Transport of VAs was mainly in the dissolved phase, resulting in limited load reductions by buffers. To achieve meaningful VA load reductions at field‐scale requires a source‐to‐buffer area ratio of