This forecast was produced using published data reporting:

- Capture locations of the rodent reservoir
*Mastomys natalensis* - Presence or absence of Lassa virus in
*M. natalensis* - Seasonal abundance of
*M. natalensis*occupying human habitations

The underlying model uses machine learning to fit three sub-models: the first two sub-models are described in the paper:

Basinski, AJ et al, 2020. Bridging the gap: Using reservoir ecology and human serosurveys to estimate Lassa virus incidence in West Africa. BioRxiv.

These sub-models return the combined probability that, in a pixel with
a specific set of environmental attributes, both *M. natalensis* and
the Lassa virus are present. These sub-models use boosted
classification trees developed in the R programming language, and the
corresponding code is publicly available on github
at this
link.

The third sub-model returns the probability that a human habitation is
occupied by *M. natalensis*, and is fit to rainfall patterns over the
previous year using principal components regression. Specifically, in
the third sub-model, monthly rainfall averages over a year are encoded
into z-scores of two principal components. The principal components,
in turn, describe annual rainfall patterns across West Africa. Lassa
spillover risk, as depicted in the map, is proportional to the
probability that a Lassa-infected rodent is present in a human
habitation (product of the three sub-model outputs).