Development and implementation of a passive surveillance system for Aedes albopictus in Emilia-Romagna, Italy

The invasive Asian tiger mosquito (Aedes albopictus) has become an important public health concern in Italy, particularly in the Po Valley area, where its biting behaviour and nuisance have contributed to multiple outbreaks of mosquito-borne diseases over the last two decades. To address this growing threat, the Emilia-Romagna region has conducted intensive mosquito monitoring efforts since 2010, generating a rich observational database. Taking advantage of this rich oviptrap dataset, we implemented a stacked machine learning approach to predict the distribution and abundance of Ae. albopictus, based on ovitrap data and environmental covariates from 2010 to 2023. A spatio-temporal sensitivity analysis was carried out to determine the amount of data necessary to train a reliable model. Our results revealed that models trained on fewer years of data but with broader spatial coverage statistically outperformed those trained on longer time frames. This indicates that including data from diverse environmental settings improves model performance more than simply increasing the temporal depth of the training set. Models that incorporated a larger number of sampling locations were also more effective at capturing complex environmental influences on mosquito populations. Despite underestimating peak summer abundance in 2023, the model effectively and consistently predicted the seasonal trends and the spatial distribution of Ae. albopictus. These findings highlight the potential of such models to inform public health strategies and optimise mosquito control interventions to mitigate vector-borne disease risks and nuisance, besides emphasising the importance of long-term and spatially extensive data in improving model performance.