Kiwifruit vines don't walk. They don't fly. They are rooted to their trellises on orchards scattered across New Zealand. And yet in 2010, Pseudomonas syringae pv. actinidiae — a bacterial pathogen that devastates kiwifruit — spread from one region to the next in a pattern that had nothing to do with wind, water, or proximity.
Rao, Higgs, Kingi, Radicchi, Fortunato, and Litvinova (arXiv:2602.21491) show that the epidemic followed the commute. Human mobility data — the aggregate flow of people between regions — predicts the spatiotemporal pattern of the outbreak. The pathogen hitched rides on shoes, vehicles, pruning equipment, harvest bins. It moved along roads, not through air.
The model is a metapopulation framework calibrated against real mobility data. Each orchard is a node. The connections between them are weighted by human traffic — not geographic distance, not wind corridors, not shared irrigation. The key finding is about the relative importance of local versus long-range dispersal. Most transmission events are local: neighboring orchards, shared workers, overlapping supply chains. But the nationwide spread — the transition from a local outbreak to a national crisis — depends on rare long-range jumps. A seasonal worker drives from Waikato to Bay of Plenty. A harvesting crew moves from one district to another following the season. These infrequent long-range connections seed distant outbreaks that then grow locally.
The temporal structure matters too. Outbreak severity depends critically on when the pathogen arrives relative to the agricultural calendar. Peak traffic periods — harvest season, pruning season — are peak dispersal periods. The same pathogen introduced during the quiet months might remain local. During the busy months, it rides the labor wave.
This has a practical consequence for intervention design. If spread follows human mobility rather than geographic proximity, then surveillance and quarantine strategies should target the mobility network, not the map. Monitoring orchards adjacent to an outbreak is important but insufficient. Monitoring orchards connected to the outbreak via labor flow, supply chains, and equipment sharing is what actually intercepts the pathogen before it jumps.
Plants get sick where people go. The epidemic map is a commute map.
Based on V.K. Rao, R. Higgs, H. Kingi, F. Radicchi, S. Fortunato, and M. Litvinova, "Modeling plant disease spread via high-resolution human mobility networks" (arXiv:2602.21491, February 2026).