River Basins Predict Amazon Malaria Outbreaks Better Than Political Borders

River basins can predict malaria outbreaks much better than standard political boundaries, according to a recent retrospective analysis of infection records in the Peruvian Amazon. When scientists mapped ten years of disease surveillance data across natural watersheds rather than government districts, their computer models cut prediction errors by more than half.

Malaria continues to pose a persistent public health challenge across tropical South America, where transmission tends to cluster tightly in forested ecoregions. In Peru, more than 95 percent of all reported infections concentrate in the northeastern department of Loreto. The disease is primarily driven by Plasmodium vivax and Plasmodium falciparum parasites, which rely on environmental conditions like temperature, humidity, and rainfall to complete their life cycles and multiply within mosquito populations.

Traditionally, health agencies track and forecast outbreaks using administrative boundaries such as districts and provinces. While these political units are useful for logistics, they rarely reflect the micro-environmental factors that govern mosquito breeding and human movement. To test whether ecological boundaries offer a sharper lens, researchers evaluated monthly geocoded malaria cases from 1,829 villages between 2009 and 2018.

The study, published in PLOS Global Public Health, compared traditional districts against three different scales of watershed units, known as hydrobasins, alongside official boundaries set by Peru’s National Water Authority. The team incorporated satellite-derived meteorological and environmental metrics—including soil moisture, runoff, land surface temperature, and vegetation indexes—into a series of statistical models and machine learning algorithms.

Why Watersheds Outperform Political Borders

The choice of spatial unit changed predictive accuracy materially. Across every tested algorithm, hydrobasin level 7 units consistently achieved the lowest root-mean-square error. For Plasmodium falciparum, models built on level 7 river basins returned a prediction error of 28.67, which was roughly 35 percent lower than districts and nearly 70 percent lower than coarser hydrobasin divisions. A similar advantage appeared for Plasmodium vivax, where level 7 catchments yielded an error of 73.85 compared to more than 200 for alternative spatial divisions.

Natural drainage divides shape the terrestrial and aquatic ecosystems where mosquitoes find standing water to lay their eggs. Furthermore, rural populations in the Amazon travel almost exclusively along river networks for work and daily life. Aggregating disease data within hydrobasins captures these shared hydrological processes and human mobility pathways much more faithfully than arbitrary political lines.

Differences Between Parasite Species

The research also revealed distinct performance gaps between the two malaria species. Predictive models for Plasmodium falciparum remained stable and accurate across different validation tests. In contrast, models tracking Plasmodium vivax exhibited higher errors and greater variability. The authors suggest this discrepancy stems from biological complexities unique to Plasmodium vivax, such as dormant liver stages that trigger relapses and a broader ecological tolerance that is harder to capture using environmental covariates alone.

Despite their strong predictive performance, the watershed models displayed notable calibration tendencies. Residual analysis revealed a general tendency for algorithms to over-predict case counts in individual spatial units rather than under-predict them. Furthermore, the authors noted several limitations, including the reliance on passive surveillance data that captures only patients who visit healthcare facilities, leaving asymptomatic and submicroscopic infections uncounted.

Health authorities can integrate these hydrological frameworks into existing administrative workflows without upending current logistics. By using hydrobasin-level risk maps to prioritize sub-district interventions, officials can deploy targeted vector control measures—such as indoor residual spraying and larval management—directly to high-risk environmental corridors.

The research was published in PLOS Global Public Health on November 2, 2026.