Data-Driven Modeling to Assess Receptivity for Rift Valley Fever Virus

Abstract

Rift Valley Fever virus (RVFV) is an enzootic virus that causes extensive morbidity and mortality in domestic ruminants in Africa, and it has shown the potential to invade other areas such as the Arabian Peninsula. Here, we develop methods for linking mathematical models to real-world data that could be used for continent-scale risk assessment given adequate data on local host and vector populations. We have applied the methods to a well-studied agricultural region of California with 1 million dairy cattle, abundant and competent mosquito vectors, and a permissive climate that has enabled consistent transmission of West Nile virus and historically other arboviruses. Our results suggest that RVFV outbreaks could occur from February–November, but would progress slowly during winter–early spring or early fall and be limited spatially to areas with early increases in vector abundance. Risk was greatest in summer, when the areas at risk broadened to include most of the dairy farms in the study region, indicating the potential for considerable economic losses if an introduction were to occur. To assess the threat that RVFV poses to North America, including what-if scenarios for introduction and control strategies, models such as this one should be an integral part of the process; however, modeling must be paralleled by efforts to address the numerous remaining gaps in data and knowledge for this system. Rift Valley fever virus is a pathogen enzootic to sub-Saharan Africa, with epidemic transmission occurring sporadically between mosquitoes and mammals, notably livestock. The virus is regarded as a global threat to agriculture and human health because it has proven capable of expanding its range into western and northern Africa, Madagascar, and the Arabian Peninsula, and a recent study has shown that mosquitoes in North America are capable of transmitting the virus. Here, we used a set of mathematical equations to formulate a logical representation of potential transmission mechanisms, and we informed the model with real-world data and generalizable methods to define spatial and temporal variation in mosquito and host abundance. We applied these methods in California's warm, agricultural Central Valley, an area with a history of mosquito-borne virus transmission and a hub of California's dairy industry. Model-derived transmission estimates indicated broad potential for transient epidemics that could result in economic losses in livestock in all but the coldest winter months, but the greatest risk for intense, sustained transmission occurred during the summer when both vector abundance and temperatures were highest. We also highlight critical gaps in the data available to inform models for Rift Valley fever virus.