Self-organizing maps for analysing pest profiles: Sensitivity analysis of weights and ranks

•Species ranks are much less influenced by variations in dataset size than species weights.•SOM PPA should be suitable for studying small datasets restricted to only a few species.•Minor data pre-processing is needed before SOM PPA, resulting in little modeller-induced bias.

Self organizing maps for pest profile analysis (SOM PPA) is a quantitative filtering tool aimed to assist pest risk analysis. The main SOM PPA outputs used by risk analysts are species weights and species ranks. We investigated the sensitivity of SOM PPA to changes in input data. Variations in SOM PPA species weights and ranks were examined by creating datasets of different sizes and running numerous SOM PPA analyses. The results showed that species ranks are much less influenced by variations in dataset size than species weights. The results showed SOM PPA should be suitable for studying small datasets restricted to only a few species. Also, the results indicated that minor data pre-processing is needed before analyses, which has the dual benefits of reducing analysis time and modeller-induced bias.