This forecast was produced using data that describes:
- Presence or absence of Lassa virus in M. natalensis
- Seasonal abundance of M. natalensis occupying human habitations
The underlying model is based off of an earlier version of the forecast described in
Basinski AJ, Fichet-Calvet E, Sjodin AR, Varrelman TJ, Remien CH, Layman NC, et al. (2021)
Bridging the gap: Using reservoir ecology and human serosurveys to estimate Lassa virus
spillover in West Africa. PLoS Comput Biol 17(3): e1008811.
https://doi.org/10.1371/journal.pcbi.1008811
The forecast shown here uses machine learning to fit two sub-models that calculate the probability that both
M. natalensis and the Lassa virus are present within each 5 x 5 km2 pixel. The first submodel uses boosted
regression trees to predict the trap-success (a measure of abundance) of Mastomys natalensis rodents in
houses. These predictions are made using static environmental predictors that describe the average climate, the land
cover type, elevation, and human population density. To incorporate a temporal component of spillover risk, this model
also uses lagged measures of monthly precipitation. The second submodel uses boosted regression trees to predict the
probability that Lassa virus is circulating in the local Mastomys natalensis population. This prediction is
made using static environmental data. The Lassa risk reported in this map is the product of these two outputs.