Rainfall, Rodent Recruitment, and a Transferable Model of Zoonotic Risk
Climate signals are only useful when they can be translated into host biology. This study does exactly that by tying seasonal rainfall to rodent demography, infection dynamics, and ultimately to the timing of human Lassa fever outbreaks through a mechanistic framework built around the Natal multimammate mouse, Mastomys natalensis.
The work combines an integral projection model of rodent demography with a compartmental disease model inside a Bayesian framework. Rather than treating weather as a loose correlate of infection, the analysis connects climate inputs to host recruitment, infection prevalence, and the persistence of a related arenavirus in the rodent reservoir. In practical terms, that makes the model more than a regression exercise; it becomes a biological simulation of how environmental forcing can move through a host population and shape zoonotic hazard.
Seasonal Rainfall Drove Recruitment in the Reservoir Host
Data from 20,249 captures of Mastomys natalensis collected between 1994 and 2023 anchored the demographic side of the model, while 7,850 serological tests from 2010 to 2017 informed the infection component. Seasonal rainfall emerged as the dominant driver of rodent recruitment, explaining 52.1% of modeled recruitment variation, with a 95% credible interval of 2.1% to 79.0%.
Rainfall matters here because rodent population turnover changes the age structure of the host community. When recruitment rises after wet periods, the population is replenished with susceptible juveniles and subadults, and infection can move through a younger cohort that has not yet accumulated immune protection or behavioral separation from contaminated environments. In a reservoir system, that demographic churn can be as important as the pathogen itself.
One result stands out with unusual force. Infected pregnancies were estimated to produce vertical transmission in 79.1% of cases, with a 95% credible interval from 69.3% to 88.9%.
Vertical Transmission Appears to Sustain Viral Persistence Between Seasons
That estimate makes vertical transmission the primary mechanism sustaining viral persistence across breeding seasons. For a rodent-borne arenavirus, transplacental or perinatal passage gives the pathogen a direct route across gaps in transmission that would otherwise be created when environmental conditions shift or adult contact patterns change. The biology is straightforward: if infected offspring enter the population already carrying virus, the system does not need constant horizontal transmission to remain endemic.
This is not a small mechanistic detail. It changes how reservoir maintenance is understood. Seasonal host abundance can rise and fall, but a virus that repeatedly crosses the placenta can exploit the host’s own reproductive cycle to bridge the leaner months. The model therefore identifies a route of persistence that is both biologically plausible and epidemiologically consequential, especially in settings where rainfall pulses structure breeding activity.
Testing Transferability Against Lassa Fever in Nigeria
The real stress test was whether the same climate-host framework could anticipate human outbreak timing elsewhere without parameter refitting. To examine that question, the model was applied to 6,469 laboratory-confirmed Lassa fever cases in Nigeria from 2018 to 2025, using climate inputs from the new setting while keeping the model structure intact.
Predicted peaks in rodent subadult infections preceded observed outbreaks by 0.92 months, with a 95% credible interval from -2.76 to 0.92. In addition, 83% of predicted and observed peaks fell within ±28 days, with Pr(Δ ≤ 28d) = 0.83. Those estimates were far better than either seasonal rainfall alone or rodent demography alone, both of which performed poorly by comparison, with Pr(Δ ≤ 28d) ≤ 0.11.
The temporal alignment is the central clinical and public health message. Subadult infection peaks tracked outbreak timing closely enough to suggest that the earliest waves of zoonotic hazard emerge before the full human case surge becomes visible in surveillance systems.
Subadults, Not Adults, Seem Closest to the Spillover Signal
Infected adult peaks lagged observed outbreaks by 0.92 months, with Pr(Δ ≤ 28d) = 0.56. That weaker temporal fit points away from adults as the primary near-term driver of spillover and toward subadults, which may be more numerous, more mobile, or more likely to interact with contaminated food stores and household perimeters during the phases of population turnover that follow rainfall-driven recruitment.
This distinction matters because it links a specific host age class to human risk. If subadults are the best temporal proxy for imminent outbreak activity, surveillance strategies built around broad rodent abundance alone may miss the more relevant biological signal. A seasonal rise in rodents is not automatically a rise in hazard; the age structure of the reservoir and the infection status of new recruits may be more informative than raw density.
Outbreak magnitude was not reproduced by the model, and the explanation is likely outside the reservoir system itself. Human behavior, household exposure, land use, food storage practices, and surveillance intensity can all magnify or mute the number of clinically detected cases once spillover has begun. A model built on host ecology can estimate when risk rises; it cannot fully determine how many patients reach care or how widely infection spreads after the first spillover events.
A Mechanistic Template for Other Climate-Sensitive Reservoir Systems
The value of the framework lies in transferability. By combining climate inputs, rodent demography, and arenavirus transmission in one mechanistic structure, the analysis provides a template that can be adapted to other zoonotic systems where weather shapes host reproduction and pathogen persistence. The approach is especially relevant for pathogens maintained in wildlife reservoirs whose population structure changes rapidly with rainfall, food availability, and seasonal breeding cycles.
For clinicians and public health teams working in endemic settings, the practical implication is not a broad forecast of every outbreak year. It is a more biologically grounded way to identify windows of elevated hazard, especially when reservoir subadult infection is rising and vertical transmission is preserving virus across breeding seasons. The model’s strength is temporal anticipation, not prediction of case counts.
The remaining limitation is equally concrete: outbreak magnitude was not captured, and the framework cannot substitute for human exposure data, household-level behavior, or surveillance completeness. Those missing elements are precisely what determine whether a rise in zoonotic risk becomes a measured outbreak, a localized cluster, or a larger public health event.