Zoonotic spillover at agricultural frontiers: risk maps exist, but prevention rarely follows

A systematic review of 59 studies has identified a major weakness in current zoonotic spillover prevention: land-use change can increasingly be detected and risk hotspots mapped, but these tools are rarely connected to real-world interventions. The largest gap lies between predicting where spillover could occur and demonstrating that acting on those predictions actually prevents infection.

The review focused on agricultural frontiers (areas where farming, pasture or plantations meet or replace natural habitats) and assessed research published between 2015 and 2026 on three linked steps: detection, prediction and prevention.

Of the 59 studies included, 31 addressed detection, 28 prediction and only 14 prevention. Most importantly, 45 studies examined just one component of the chain, and none integrated all three.

Land-use change is detected, but mostly retrospectively

Nearly all detection studies (29 of 31) relied on existing land-cover datasets such as MapBiomas, PRODES or global forest-change products. The review found no study reporting routine operational use of these geospatial tools to trigger zoonotic surveillance. Near-real-time approaches such as radar, drones or rapid clearing alerts were also absent from the reviewed evidence.

Deforestation was the most frequently studied land-use change, followed by crop expansion, pasture or livestock expansion and forest fragmentation.

Risk maps are common, but poorly transferred

Predictive mapping was better developed, but validation remained limited. Among 28 predictive studies, only 10 used independent validation, and just three tested whether models transferred successfully to a new place or a later time period. Eight studies reported no validation at all.

Yet 24 of the 28 predictive studies proposed a prevention or control use for their maps, including vaccination, vector control, farm biosecurity and land-use planning.

The critical gap: no map-targeted intervention was evaluated

This is the review’s strongest finding. None of the 59 studies evaluated whether an intervention allocated according to a risk map actually reduced spillover. In other words, prediction frequently leads to recommendations, but rarely to prospective field testing.

Prevention studies were also geographically skewed: eight of 14 were conducted in high-income countries, and measured outcomes generally stopped at livestock infection rather than human spillover.

The authors argue that future work should move beyond producing ever more sophisticated maps and instead test whether risk-based allocation of surveillance, vaccination, biosecurity or vector control performs better than conventional approaches.

So, risk mapping should not be treated as prevention until it has been validated in new settings and linked to interventions whose impact is actually measured.