Validating a rapid algorithmic weed hazard ranking method

<b>Background: </b>Conventional weed risk assessments (WRAs) are time-consuming and often constrained by species-specific attribute data gaps. We present a validated, algorithmic alternative, the <b><i>CPG</i></b> model, that integrates climatic suitability (<b><i>C</i></b>), weedrelated publication frequency (<i>P</i>) and global occurrence data (<b><i>G</i></b>), using publicly available databases and artificial intelligence (AI)-assisted text screening with a large language model (LLM).<b>Results</b>: The <b><i>CPG</i></b> model was tested against independent weed hazard classifications for New Zealand and California. In New Zealand, 89% of 480 randomly selected plant taxa had sufficient data to generate scores, which aligned well with expert classifications and moderately with outputs from the 48-question WRA [Pheloung PC, Williams PA, and Halloy SR. <i>Journal of </i><i>Environmental Management</i> 57:239–251 (1999)]. For more than 5000 species assessed using Randall's 14-criteria generalised risk model [Randall RP, 20th Australasian Weeds Conference: 5–12 (2016)], all <b><i>CPG</i></b> variables were informative. The model also showed strong agreement with a 19-criteria California weed hazard system. Multinomial regression and receiver operator characteristic curve (ROC) analyses confirmed consistent predictive performance, with true-positive rates from 0.69 to 0.90 and truenegative rates from 0.71 to 0.97. It effectively distinguished high- from low-hazard species. Sensitivity analysis showed that as evidence for weediness increased, score stability improved, supporting robust rankings for high-hazard species.<b>Conclusion</b>: The <b><i>CPG</i></b> model offers a transparent, scalable and cost-effective tool for early-stage weed hazard screening. It delivers substantial time savings over attribute-based WRAs while maintaining alignment with expert evaluations. The model enables rapid triage of large species lists, including actual or potentially introduced taxa under current and future climates, supporting prioritisation for detailed risk or management feasibility assessments. Its automation and reproducibility make it a valuable tool for global biosecurity and invasive species management.