Towards pest outbreak predictions: Are models supported by field monitoring the new hope?

Physiologically-based models are the core of Decision Support Systems (DSS) for insect pest and disease control in cultivated fields. However, the large-scale use of DSS remains scarce and limited, despite the continuous update and formulation of new models by the literature. The main reason behind this lack of real-world use relates to the purely descriptive approach of these models, which are usually validated a posteriori. The major limiting factors that preclude the use of these tools for prediction purposes are their dependence on time zero and initial abundance to start the simulations. In this study, we present a theoretical framework that includes field monitoring data as an active part of a pest population density model simulation, which helps to overcome these obstacles. More specifically, we propose the application of an estimator scheme in the form of an Extended Kalman Filter (EKF) to a revised physiologically-based model from the literature. In the paper, we carry out a preliminary test of the theoretical framework applied to the case of Drosophila suzukii. This case study shows that the dependence of the simulations on the initial conditions and time zero is strongly reduced by using the EKF. Overall, the outcome of this research indicates that an estimator scheme is a necessary step to move from description to prediction in the pest population modelling field. •Application of an Extended Kalman Filter (EKF) in pest population models.•We revised a physiologically-based model to apply it to an EKF subject to different pest monitoring techniques.•The framework was tested in the case of Drosophila suzukii.•The EKF limits the dependence of the model solution on initial conditions and substantially improves the model estimation.