Problems in qualifying and quantifying assumptions in plant protection models: Resultant simulations can be mistaken by a factor of million

Models related to plant protection can give an appreciation of a phenomenon, and provide ideas for priorities in research, as well as suggest management strategies. The problem with models is that mistakes can be huge when inaccurate assumptions are made about key parameters, as described with three sets of models: (1) our own model predicting that five herbicide-resistant Striga plants would appear per hectare per season was based on an inaccurate assumption that heterozygotes would be selected, and a heterozygous mutation frequency was used, while a recessive mutant frequency should have been used. A revised model with a recessive mutation would predict five resistant plants per million hectares per season; (2) the model predicting that Bt resistant insects would quickly evolve in transgenic cotton and maize unless massive refuges were instituted, assuming a single binding site for the toxin and minor unfitness of resistant individuals, not realizing that resistant individuals may be extremely unfit or that Bt may have multiple targets; (3) a model that claims that unfit transgenes from crops would decimate wild relatives by swamping. The model assumed animal-type low progeny numbers, and did not consider competition for replacement, nor the infrequency that crop pollen could reach wild relatives. Models must be retrospectively critiqued, based on field data and new knowledge, and not be allowed to become accepted as being axiomatic.