Disentangling the role of environmental and human pressures on biological invasions across Europe

Abstract

The accelerating rates of international trade, travel, and transport in the latter half of the twentieth century have led to the progressive mixing of biota from across the world and the number of species introduced to new regions continues to increase. The importance of biogeographic, climatic, economic, and demographic factors as drivers of this trend is increasingly being realized but as yet there is no consensus regarding their relative importance. Whereas little may be done to mitigate the effects of geography and climate on invasions, a wider range of options may exist to moderate the impacts of economic and demographic drivers. Here we use the most recent data available from Europe to partition between macroecological, economic, and demographic variables the variation in alien species richness of bryophytes, fungi, vascular plants, terrestrial insects, aquatic invertebrates, fish, amphibians, reptiles, birds, and mammals. Only national wealth and human population density were statistically significant predictors in the majority of models when analyzed jointly with climate, geography, and land cover. The economic and demographic variables reflect the intensity of human activities and integrate the effect of factors that directly determine the outcome of invasion such as propagule pressure, pathways of introduction, eutrophication, and the intensity of anthropogenic disturbance. The strong influence of economic and demographic variables on the levels of invasion by alien species demonstrates that future solutions to the problem of biological invasions at a national scale lie in mitigating the negative environmental consequences of human activities that generate wealth and by promoting more sustainable population growth.

Fig. 1.

Partitioning of the variance of standardized alien species numbers in each region adjusted for the effect of area among classes of explanatory variables representing insularity (as the only fitted geographic variable), human factors (including wealth and human population density) and climate. Model for all taxa and for individual taxonomic groups, ranked according to explained variance, are presented. Only the classes of variables for which their joint effects appear significant (P < 0.05) are shown (i.e., three classes of variables for all taxa, bryophytes, plants and reptiles, and two classes for birds, mammals and aquatic invertebrates). Intersections of circles in the Venn diagrams represent the variance jointly explained by two or three classes of variables; their negative fractions are given by the correlation structure of the explanatory variables, in which direct and indirect positive and negative effects of the individual variables are combined (ref. , p. 533). The rectangles represent 100% of the variance, of which [h] (for taxa having three classes of variables with significant joint effects) or [d] (for taxa having two such classes) is the unexplained part. Only the net effects ([a], [b] and [c] for taxa in which three classes of variables have significant joint effects, and [a] and [c] for taxa with two such classes) can be tested statistically; of these, significant (P < 0.05) net effects are underlined. Intersections ([d]–[g] for taxa with three classes and [b] for taxa with two classes) cannot be tested. Note that there is no variation partitioning for insects and fungi because all variance in these taxa was explained by the explanatory variables from only human factors (29.5% for insects and 24.8% for fungi) and that no explanatory variable appeared significant for amphibians and fish.

Fig. 2.

Regression tree analysis of standardized alien species numbers in each region adjusted for the effect of area for all taxa. Each node of the tree is described by the splitting variable and its split value (population density in inhabitants/km², wealth in US$ per capita, mean and SD of standardized alien species numbers adjusted for area, and the number of samples at that node). Vertical depth of each node is proportional to its improvement value that corresponds to explained variance at each node. The total variance explained is R² = 0.256. (Inset) Cross-validation processes for selection of the best regression trees. Line shows a single representative 10-fold cross-validation of the most frequent (modal) best trees with SE estimates of each tree size. Bar charts are the numbers of the optimal trees of each size (frequency of tree) selected from a series of 50 cross-validations based on the minimum cost tree, which minimizes the cross-validated error (white, SE rule 0), and 50 cross-validations based on the one-SE rule (gray, SE rule 1), which minimizes the cross-validated error within one SE of the minimum. The most frequent (modal) trees have four terminal nodes. See Table S1 for region codes.