Fig. 1

The workflow of the study. We collected literature data (a) in the period 2010–2017 about the prevalence of Borrelia burgdorferi s.l. (Bb) and B. miyamotoi in questing ticks (but only data from questing nymphs were finally used). This allowed to plot raw maps of the reported distribution of each species of Bb. We built scripts in R to mass download 8-days 1 km. satellite images from MODIS repository (b) that were transformed into Fourier components using in-house developed scripts (c). All the scripts and the transformed satellite imagery are available in http://datadryad.org/resource/doi:10.5061/dryad.2h3f2.2. We tessellated the complete territory with a grid of hexagonal cells of 0.25° of radius, and environmental data as well as prevalence values (converted to categories) were transferred (d). The categories of prevalence for each species were also transferred (e) to a standard definition of the ecological classes in Europe (LANMAP2) to obtain wide area estimates of Bb species composition and prevalence. Data on prevalence and environmental traits were used to produce a map in the environmental niche about the restriction of Bb species and prevalence to different portions of the niche outlined by temperature and vegetation (f). These grid-derived values were also used to model the influence of the environmental traits on the observed values of prevalence by training a Neural Network