Machine Learning

An overwhelming absence of information on areas risk toward invasive species introduction results out of the significant time and labour requirements of species-specific analyses which complicates management. We aimed to develop a generic modelling approach to identify hotspots for plant pest introductions. We assessed the risk of presence in Europe for the whole group of 248 invasive species on the priority lists (A1 and A2) of the European and Mediterranean Plant Protection Organization. Global georeferenced data on a wide range of potential predictors related to climate, soils, water, and anthropogenic factors were collected, and an elastic-net machine learning algorithm was trained on around 341 000 observations across the globe to predict new introduction of invasive species as a function of the predictors. The algorithm was tuned and trained for nine setups resulting from the combinations of three approaches to generating background data and three cross-validation techniques.

Xylella fastidiosa is one of the most dangerous plant-pathogenic bacteria worldwide. Regulatory measures were enacted in response to the detection of the subsp. pauca (Xfp) in Italian olives in 2013, but the current impact is nevertheless major. We developed a spatially explicit bio-economic model to compute potential future economic impact of the Xfp strain. Uncertainty on spread is accounted for by simulating different scenarios. The majority of orchards were found to be within climatically suitable territory. Even under slow disease spread and the ability to replant with resistant cultivars, projections of future economic impact in affected countries run in the billions of Euros. Our findings highlight the importance of minimizing disease spread and implementing adaptation measures in affected areas.