Sampling bias mitigation for species occurrence modeling using machine learning methods

Forecasting Outbreaks of Hantaviral Disease: Future Directions in Geospatial Modeling

Hantaviral diseases have been recognized as 'place diseases' from their earliest identification and, epidemiologically, are tied to single host species with transmission occurring from infectious hosts to humans. As such, human populations are most at risk when they are in physical proximity to suitable habitats for reservoir populations, when numbers of infectious hosts are greatest. Because of the lags between improving habitat conditions and increasing infectious host abundance and spillover to humans, it should be possible to anticipate (forecast) where and when outbreaks will most likely occur. Most mammalian hosts are associated with specific habitat requirements, so identifying these habitats and the ecological drivers that impact population growth and the dispersal of viral hosts should be markers of the increased risk for disease outbreaks. These regions could be targeted for public health and medical education. This paper outlines the rationale for forecasting zoonotic outbreaks, and the information that needs to be clarified at various levels of biological organization to make the forecasting of orthohantaviruses successful. Major challenges reflect the transdisciplinary nature of forecasting zoonoses, with needs to better understand the implications of the data collected, how collections are designed, and how chosen methods impact the interpretation of results.