Weed Seeds in New Zealand Vegetable Seed Lots: Evaluating Seed Cleaning Effectiveness From Commercial and Biosecurity Perspectives

Weed seed contamination in vegetable seed lots poses risks to commercial seed quality and biosecurity, yet few studies have examined contamination in vegetable crops, and none have compared pre‐ and post‐cleaning levels using quantitative seed counts. We analysed 1,322 field‐dressed (FD) and machine‐dressed (MD) purity test results from four vegetable seed crops (beet, brassica, radish, spinach) grown in New Zealand between 2016 and 2020, reporting 91 contaminant taxa across 72 genera. A Bayesian negative binomial generalised linear mixed model identified contaminants most likely to persist after cleaning and assessed covariate effects. Cleaning was effective: Approximately half of contaminated seed lots had all contaminants removed, and median seed counts were reduced by 55–82% across crops. Complete contaminant removal varied from 8% of beet lots to 59% of brassica lots, and some contaminants persisted in high numbers, with up to 200 weed seeds in a single MD sample. Galium aparine was predicted to be the most persistent contaminant across all crops, followed by Malva parviflora , driven by high field prevalence and morphological similarity to crop seed that limited the effectiveness of cleaning. Additional persistent contaminants included other crop species, with volunteer cereals particularly persistent in spinach. Chenopodium album and Polygonum aviculare were among the most effectively removed, although C. album remained persistent in brassica and spinach due to high initial FD levels. Pre‐cleaning contamination was the strongest predictor of post‐cleaning counts, and open‐pollinated spinach contained substantially more weed seeds than hybrid lots. No regulated quarantine weeds, environmental weeds or species new to New Zealand were reported in MD samples, indicating contamination posed a commercial rather than biosecurity risk, though some contaminants identified only to genus level belong to genera containing regulated species. This study provides a novel quantitative seed cleaning analysis and a transferable framework for evaluating treatment effectiveness in zero‐heavy biosecurity count data.