Spotting Virus from Satellites: Modeling the Circulation of West Nile Virus Through Graph Neural Networks

The occurrence of West Nile Virus (WNV) represents one the most common mosquito-borne zoonosis viral infections. Its circulation is usually associated with climatic and environmental conditions suitable for vector proliferation virus replication. On top that, several statistical models have been developed to shape forecast WNV circulation: in particular, recent massive availability Earth Observation (EO) data, coupled continuous advances field Artificial Intelligence, offer valuable opportunities. In this paper, we seek predict by feeding Deep Neural Networks (DNNs) satellite images, which extensively shown hold features. Notably, while previous approaches analyze each geographical site independently, propose a spatial-aware approach that considers also characteristics close sites. Specifically, build upon Graph (GNN) aggregate features from neighbouring places, further extend these modules consider multiple relations, such as difference temperature soil moisture between two sites, well distance. Moreover, inject time-related information directly into model take account seasonality spread. We design an experimental setting combines images – Landsat Sentinel missions ground truth observations Italy. show our proposed Multi-Adjacency Attention Network (MAGAT) consistently leads higher performance when paired appropriate pre-training stage. Finally, assess importance component MAGAT ablation studies.