Modeling plant disease spread via high-resolution human mobility networks

Abstract

Human mobility plays a crucial role in the spread of human diseases, but is rarely quantified in plant disease epidemics. To address this gap, we integrate a unique, high-resolution network of human movements in New Zealand with a metapopulation model to mechanistically simulate pathogen transmission. We calibrate the model on the nationwide 2010 kiwifruit vine disease (Psa-V) outbreak, and show that it accurately reproduces the observed spatiotemporal spread, confirming that the human mobility network is a strong foundation for modeling transmission dynamics. By analyzing spatial infection trends, we find that most dispersal occurs locally, as often illustrated in the plant-outbreak literature. However, sporadic long-range connections are necessary to model a nationwide outbreak. Using the model as an in-silico laboratory, we demonstrate that enhanced surveillance accelerates detection and that outbreak severity is highly sensitive to the timing and location of initial disease importation. We observe a potential causal link between seasonal labor patterns and epidemic risk in high-traffic seasons. This study provides a robust, data-driven framework for modeling and predicting the spatiotemporal spread of agricultural pathogens. It underscores the importance of leveraging human mobility networks to design timely interventions and surveillance systems, protecting global food security.

Figures (7)

• Figure 1: Map of the horticulture industry in New Zealand. Each point represents a property in the Onside dataset ($N=2,281$ total properties). Black points denote the $S=1,300$ properties located in the Bay of Plenty, which are used as seeds for the spreading dynamics simulated by our model.
• Figure 2: Schematic representation of mobility networks. Using the Onside movement dataset, we generate networks representing movements of workers between properties. In all networks, we measure the number of movements of people between properties, which vary based on the source and destination. A movement is directed in nature and is recorded when a person leaves a source property and checks into a destination property. In this figure, the size of the tractor corresponds to the number of movements: the larger the tractor, the more movements occur. Depending on the aggregation level of the data, we define either yearly, seasonal, or monthly networks. People moving between properties spread plant disease, and a higher frequency of movements between properties increases the probability of spreading. We note that starting the infection during different seasons or months can lead to different epidemic outcomes, since monthly movement networks differ. Infections are colored red showing how seeding infections during different seasons or months can lead to different epidemic pathways.
• Figure 3: Dynamics of the Psa-V epidemic in New Zealand. (A) Number of new monthly infected properties and (B) cumulative number of monthly infected properties. The solid curve in both panels represents the raw data extracted from the report written by Greer and Saunders greer2012costs. In this, the monthly infection data starts from February 2011 and ends in March 21st, 2012, but the actual outbreak started in November 2010 and has now become endemic in New Zealand. Thus, the leftmost point in the raw data corresponds to the number of properties that became infected between November 2010 and February 2011. The gray curve is a 2-month moving average of the raw timeseries. The moving average is calculated starting from March 2011.
• Figure 4: Best-fitting epidemic models. (A) We plot the incidence timeseries obtained from the epidemic model when setting its parameter values to their corresponding best estimates. Different colors refer to different levels of temporal aggregation of the movements between properties: yearly (blue), seasonal (orange), and monthly (green). The shaded areas define the uncertainty associated with each of these curves, computed by taking the minimum and maximum values of the incidence measured over the viable seeds for the best-fitting values for $\beta_b,\beta_w,D$ . Model's curves are compared against raw data (black). The inset displays the percentage of viable configurations for the parameters in violet, as well as the percentage of unique seeds appearing in the best-fitting configuration when holding $\beta_b,\beta_w,D$ to their best fitting value. This analysis was conducted for the various levels of temporal aggregation that can be in the model. (B) Same as in (A), but obtained when fitting the model against the 2-month moving average incidence curve (gray dashed).
• Figure 5: Spatio-temporal characterization of the spreading. Two distance metrics are compared using three different statistics: the mean, minimum, and maximum. The detection-distance metric is calculated by comparing the distances of all properties that are infected within the same month. The contact distance is calculated by finding the average distance of a newly infected property’s contacts. A contact is defined as a previously infected property that has an outgoing link to the newly infected property. We define a newly infected property as one that is detected as infected, not simply if it can transmit infection.