Modeling an introduction of invasive alien species through container cargo considering the logistics network
Abstract
Invasive alien species introduced through trade cargo outside their natural range are recognized as one of the main causes of biodiversity destruction. To prevent the invasion of these species, it is crucial to accurately estimate the relative risk of introduction at ports during the early stages of invasion. Statistical models provide essential tools for predicting this probability.
The risk of introduction is affected by the multiple locations through which the cargo has passed. Despite a large number of models stochastically explaining the expansion of the species distribution proposed, methods have not been established to consider a logistics network in focusing on the introduction stage. Spatial statistical models have been mainly used in ecology and epidemiology to estimate the probability of rare events by assuming spatial correlation and using neighboring observational data. For the species introduction through container cargo observed in multiple ports, a correlation based on the intensity of the logistics connectivity would be assumed rather than a correlation with geographical distance.
This study applied the spatial statistical models, particularly the conditional autoregression model, to the introduction of alien species to consider the logistics network between ports. The survey data on non-native ants obtained at Japanese ports were used for actual data. This species has been observed frequently in Japanese ports recently and is currently in the introduction stage. A hierarchical Bayesian model was implemented, considering the volume of imported containers to Japanese ports and the domestic container cargo network. The proposed model was compared with the model that did not consider the network based on the deviance information criterion. The result suggests that the spatial random effect based on the logistic network can capture the uncertainty of human-mediated species dispersion. We also outline the characteristics of estimated relative introduction risk in Japanese ports.