BACKGROUND : Ectoparasites exhibit pronounced variation in life history characteristics such as time spent on the
host and host range. Since contemporary species distribution (SD) modelling does not account for differences in
life history, the accuracy of predictions of current and future species’ ranges could differ significantly between life
RESULTS : SD model performance was compared between 21 flea species that differ in microhabitat preferences and
level of host specificity. Distribution models generally performed well, with no significant differences in model
performance based on either microhabitat preferences or host specificity. However, the relative importance of
predictor variables was significantly related to host specificity, with the distribution of host-opportunistic fleas
strongly limited by thermal conditions and host-specific fleas more associated with conditions that restrict their
hosts’ distribution. The importance of temperature was even more pronounced when considering microhabitat
preference, with the distribution of fur fleas being strongly limited by thermal conditions and nest fleas more
associated with variables that affect microclimatic conditions in the host nest.
CONCLUSIONS : Contemporary SD modelling, that includes climate and landscape variables, is a valuable tool to study
the biogeography and future distributions of fleas and other parasites taxa. However, consideration of life history
characteristics is cautioned as species may be differentially sensitive to environmental conditions.
Additional file 1: Calculation of coefficients of harmonic regression for
climate variables and topography in R. Description of data: R script
illustrating the production of the coefficients of harmonic regression for
climate variables and topography.
Additional file 2: Collinearity among the 19 final individual predictor
variables. All final predictor variables had correlation values below 0.7 or
above−0.7. Description of data: The table contains collinearity among
final 19 predictor variables chosen for modelling.
Additional file 3: List of predictor variables indicating their respective
reference codes. Description of data: The table contains a list of all
predictor variables considered for this study and their abbreviations as
Additional file 4: Calculation of the true skill statistic (TSS) for each
replicate of the 10-fold cross-validation from MaxEnt output in R. Description
of data: R script illustrating the calculation of the true skill statistic (TSS) for
each replicate of the 10-fold cross-validation from MaxEnt output.
Additional file 5: Variable importance (i.e. percent relative predictor
variable individual contribution) in MaxEnt models, averaged across flea
species based on (a) microhabitat preference and (b) host specificity (see
Additional file 2 for variable reference code). Significant differences in the
contribution of predictor variables between the two categories of species
are indicated by asterisks: *** P < 0.001, ** P < 0.01, * P < 0.05. Description of
data: The bar plot figure illustrate variable importance (i.e. percent relative
predictor variable individual contribution) in MaxEnt models, averaged
across flea species based on microhabitat preference and host specificity.