Source: GeoHealth
West Nile virus is the most common mosquito-borne illness in the continental United States and can in rare cases lead to a much more serious disease with an approximately 10% fatality rate. West Nile virus neuroinvasive disease (WNND) has resulted in around 3,000 deaths since its introduction to the country in 1999, but to date no national forecast for the disease exists.
Harp et al. developed a climate-informed, regionally determined forecast method for WNND cases across the United States that outperforms current benchmarks. Key to their success was aggregating historically low county-level caseloads to the regional level, the authors say. Their work highlights key climatic factors and how their regional variation affects WNND rates.
Both mosquitoes and passerine birds (a group that includes more than half of all bird species) are vectors for West Nile virus, meaning caseloads are contingent on the environmental factors affecting these species. The authors picked the most relevant climatic factors as model inputs for each region. They found that drought and temperature are most strongly linked to WNND cases overall, and precipitation is linked in some regions. The central United States saw the most consistent correlation with drought and WNND cases, whereas the northern parts of the country saw the strongest link between WNND and warmer winter and spring temperatures.
The authors compared their climate-driven model with previous benchmark models, including a simple historical caseload model and an ensemble model from a 2022 competition. They found their model consistently outperformed others across regions. Nationally, a version of their model that included both primary and secondary climate factors (such as temperature and soil moisture) offered a prediction improvement of 21.8% over the historical model.
While the advancement represents a building block toward operational West Nile virus forecasts, the authors recommend that future work focus on enhancing county-level forecasting, which would provide authorities with more actionable information to prepare for fluctuations in WNND caseloads. Future WNND forecast models may also need to overcome the issue of climate data latency to offer real-time predictions, the authors say. One option could be to incorporate weather and climate forecasts into modeling, allowing disease forecasts to look further ahead. (GeoHealth, https://doi.org/10.1029/2025GH001657, 2026)
—Nathaniel Scharping (@nathanielscharp), Science Writer
