1. BoL (TRL: 7) specializes in predicting the presence of larvae on breeding site level in rural, peri-urban and wetland breeding sites, with a particular emphasis on Culex spp. and Aedes spp. (in mediterranean areas caspius prevails). Its design incorporates various input variables, including meteorological data, Earth Observation (EO) data, which comprises meteorological conditions, derived indices pertinent to water, wetness, and land use/cover, and direct field observation data from breeding sites. The unique combination of a Variational Autoencoder and a Random Forest classifier ensures high predictive accuracy for a lead time of 7 days. The predictions of BoL have operational use for mosquito control applications in the regions of Central Macedonia and Western Greece and they also serve as a nested model for BAd.

BAd (TRL: 7) specializes in predicting the abundance of adult mosquitoes, especially Culex spp. and Aedes spp. (focusing on the nuisant floodwater mosquito caspius), on a settlement level. It integrates the predictions from BoL, EO data, meteorological data and geomorphological information to forecast mosquito abundance daily for a five-day period ahead. Lightgbm, a gradient boosting framework, forms the backbone of this model. It has been operational in the regions of Central Macedonia, Western Greece, Thessaly and Crete since 2020.

BAR (TRL: 7) is tailored for assessing the risk of West Nile Virus (WNV) transmission to humans on a settlement level. It embeds predictions from multiple sources, including sentinel birds and infected mosquitoes, and leverages both Variational Autoencoder and Lightgbm for enhanced accuracy. It has been operational in the region of Central Macedonia since 2020 providing a monthly prediction of risk for WNV transmission to humans on settlement level.