Longitudinal data analysis of meteorological variables for modeling West Nile virus in the Piedmont region

This thesis proposes a hybrid approach that combines both statistical and mechanistic models to study mosquito population dynamics in the context of the West Nile Virus (WNV) in the Piedmont region. The research extends existing entomological models by incorporating humidity as a critical factor. The study employs the GreatNector workflow in two phases. In the first phase, the CONNECTOR framework—a data-driven approach that leverages statistical methods for Functional Data Analysis (FDA)—is used to analyze longitudinal weather data. Weather characteristics, such as temperature and humidity, are integrated into the model. For each weather measurement and mosquito capture data, CONNECTOR identifies clusters of traps with similar behavior. These clusters are then combined using k-means clustering, with each cluster represented by an average trap, referred to as a “meta-trap”. In the second phase, the GreatMod framework, based on the Petri Net formalism, is used to calibrate model parameters. The meta-trap curves serve as input for this calibration, which is initially performed on one year of data. The calibrated parameters are then tested against the meta-traps from other years with similar capture behavior. This combined modeling approach provides a comprehensive framework for predicting mosquito population trends and supporting WNV containment efforts in the Piedmont region.