Darwin's naturalization hypothesis does not explain the spread of nonnative weed species naturalized in México

Despite numerous tests of Darwin's naturalization hypothesis (DNH) evidence for its support or rejection is still contradictory. We tested a DNH derived prediction stating that nonnative species (NNS) without native congeneric relatives (NCR) will spread to a greater number of localities than species with close relatives in the new range. This test controlled the effect of residence time (Rt) on the spread of NNS and used naturalized species beyond their lag phase to avoid the effect of stochastic events in the establishment and the lag phases that could obscure the NCR effects on NNS.

We compared the number of localities (spread) occupied by NNS with and without NCR using 13,977 herbarium records for 305 NNS of weeds. We regressed the number of localities occupied by NNS versus Rt to determine the effect of time on the spread of NNS. Then, we selected the species with Rt greater than the expected span of the lag phase, whose residuals were above and below the regression confidence limits; these NNS were classified as widespread (those occupying more localities than expected by Rt) and limited-spread (those occupying fewer localities than expected). These sets were again subclassified into two groups: NNS with and without NCR at the genus level. The number of NNS with and without NCR was compared using χ² tests and Spearman correlations between the residuals and the number of relatives. Then, we grouped the NNS using 34 biological attributes and five usages to identify the groups' possible associations with spread and to test DNH. To identify species groups, we performed a nonmetric multidimensional scaling (NMDS) analysis and evaluated the influences of the number of relatives, localities, herbarium specimens, Rt, and residuals of regression. The Spearman correlation and the Mann-Whitney U test were used to determine if the DNH prediction was met. Additionally, we used the clustering objects on subsets of attributes (COSA) method to identify possible syndromes (sets of biological attributes and usages) associated to four groups of NNS useful to test DNH (those with and without NCR and those in more and fewer localities than expected by Rt).

Residence time explained 33% of the variation in localities occupied by nonnative trees and shrubs and 46% of the variation for herbs and subshrubs. The residuals of the regression for NNS were not associated with the number or presence of NCR. In each of the NMDS groups, the number of localities occupied by NNS with and without NCR did not significantly differ. The COSA analysis detected that only NNS with NCR in more and fewer localities than expected share biological attributes and usages, but they differ in their relative importance.

Our results suggest that DNH does not explain the spread of naturalized species in a highly heterogeneous country. Thus, the presence of NCR is not a useful characteristic in risk analyses for naturalized NNS.