On the Estimation of Pollen Density on Non-target Lepidoptera Food Plant Leaves in Bt-Maize Exposure Models: Open Problems and Possible Neural Network-Based Solutions

摘要

Sometimes, mathematical modelling in ecology requires approximation assumptions, on the model at hand, in order to meet domain constraints, thus making the mathematical construction unrealistic. To this end, the paper analyzes a model representing the Btmaize pollen density on non-target Lepidoptera food plant leaves by a differential equation. The exact solution of the differential equation is provided, showing that the solution behavior, when the time goes to infinity, does not vanish, differently from what assumed in the model, consequently undermining the theoretical model soundness. In order to solve this drawback, the paper proposes a neuro-fuzzy model capable to obtain a robust pollen density estimate directly from data, thus avoiding unnecessary and unfeasible model approximations.