Ionizing radiation and cold plasma for post-harvest control of Sitophilus zeamais in maize: A biophysical approach with penalized logistic regression and generalised linear models

The maize weevil ( Sitophilus zeamais ) is a major pest of stored maize, causing quantitative and qualitative losses that compromise food security and trade. Chemical fumigants face increasing restrictions due to resistance and residues. Physical technologies—gamma radiation ( 60 Co), electron beam (e-beam) and cold atmospheric plasma (CAP, generated via compressed air at 7.0 W)—offer residue-free alternatives. However, standard statistical analyses (ANOVA, linear regression) are inadequate for binomial mortality data, especially when complete separation occurs (100% mortality). We evaluated the lethal efficacy of gamma rays and e-beam at doses of 0–10 kGy (10 days post-irradiation) and CAP at exposure times of 0–120 min (assessed at 20, 30 and 40 days). We employed penalized logistic regression (Firth's method) for the ionizing radiation data (to handle complete separation) and binomial GLM with probit link for the plasma dose–response. Model fit was assessed via DHARMa simulations, and multiple comparisons against control were adjusted with Dunnett's test. Both gamma rays and e-beam achieved 100% mortality at all doses ≥ 0.5 kGy ( p < 0.001). Cold plasma exhibited a time-dependent effect, reaching 98.3% mortality after 120 min ( p < 0.001 vs control). Diagnostics confirmed no overdispersion and adequate model fit. Odds ratios for e-beam and gamma were practically infinite, confirming complete lethality. The estimated LD 50 for plasma was about 12 min, and LD 99 was reached at ≈ 98 min. Very low doses (0.5 kGy) of ionizing radiation are sufficient for complete disinfestation of adult weevils. Cold plasma, though slower, offers an affordable, on-farm alternative. The use of bias-reduced logistic regression and GLM with probit provides a rigorous statistical framework for radiological entomology, avoiding the pitfalls of classical methods.