Predictive modeling of combined release strategies for Myzus persicae management: Preventive Macrolophus pygmaeus reduces aphid pressure with minimal cost increase

Abstract: We developed a hierarchical Bayesian state-space model to handle cryptic parasitoid stages and spatial variation in tritrophic greenhouse dynamics, with successful validation (100 % directional accuracy, 14 % NRMSE, ± 1 week). This framework revealed that combining specialist Aphidius colemani with preventive generalist Macrolophus pygmaeus reduces green peach aphids Myzus persicae peak density by an additional 67 % with 6 % extra cost beyond parasitoid-only programs, through temporal niche partitioning. Critically, we identified an early-phase paradox: initial monitoring significantly affects forecast accuracy yet is hardest to observe. This detectability constraint causes forecast errors to compound up to 3-fold, fundamentally limiting early-stage biocontrol forecasting when accuracy matters most.