Forecasting the spatio-temporal dynamics of Aedes Albopictus in southern Europe using a stacked machine learning model

Aedes albopictus is a competent vector of several arboviruses such as dengue, chikungunya and Zika. In the northern hemisphere, its distribution is rapidly extending northward and increasing in altitude, mediated by climate change, thermal adaptation, and the intense transportation of goods and persons. Understanding the seasonal dynamics of this species of public health concern is crucial for effective and proactive population and disease control strategies. Among the different means for monitoring Ae. albopictus populations, ovitraps are cheap and efficient tools, whose application is widespread and often used in conjunction with other mosquito control methods (e.g. BG-sentinel traps). However, resource limitations challenge public health authorities to achieve comprehensive and widespread ovitraps monitoring coverage. In this study, we assembled high-quality Ae. albopictus ovitraps observations collected from several southern European locations covering four countries (Albania, France, Italy, and Switzerland) over multiple seasons during the period 2010-2022. We used ovitraps collections and a set of environmental explanatory variables to inform a stacked machine learning model predicting the weekly average number of mosquito’s eggs that allows us to (i) infer the seasonal dynamics of Ae. albopictus over a period of 12 years, and (ii) obtain spatio-temporal explicit forecasts of mosquito egg abundance in areas not covered by traditional monitoring efforts. These results will support public health authorities' efforts for vector control, management and evaluation of the epidemic risk.