Ecological niche models and the geography of biological invasions: a review and a novel application - Invasive Plants: Ecological and Agricultural Aspects - page 38

in the US [22] and the world [35]. The results of these models qualitatively
were similar to those from our Reverse-ENM model (Figs 2C and 2D).
Mechanism to explain the disparity between the Forward- and Reverse-ENM
models
If the distribution of a species is considered representative of the niche of that
species, the Forward- and Reverse-ENM analysis suggests that niche occu-
pancy of RIFA differs between its native and invaded ranges. However, differ-
ences in predictions may be caused by mechanisms other than differences in
niche occupancy. We present several hypotheses to explain the disparity
between the Forward- and Reverse-ENM predictions, both involving mecha-
nisms that may and may not bring about differences in niche occupancy.
Hypotheses in opposition to differences in niche occupancy
Inadequate sampling in native range
To our knowledge, the native range occurrence dataset used in this study
includes all published records of RIFA in South America. However, RIFA may
actually occur south of its known native range, but it has not been observed or
collected in that region. Although the exact southern limit of RIFA is admit-
tedly not well known, inadequate sampling seems unlikely given the recent
extensive ant surveys in the area of the southern limit of RIFA, which docu-
mented ants in the genus Solenopsis with the exception of S. invicta [33]. RIFA
apparently has not been collected south of roughly 34° latitude, well north of
the Reverse-ENM predictions (Fig. 2D).
Microhabitat selection in the US
RIFA may occupy natural or disturbed microhabitats in the US that are not
indicative of the broader climate. The scale at which organisms select habitat
is finer than the resolution of the climate data used to model their distributions.
If RIFA inhabits sites that are unlike the broader climate (e.g., irrigated land),
the model may predict presence in these regions even though they are climat-
ically unsuitable. For instance, in dry climates in the US, RIFA is associated
closely with irrigated areas, such as golf courses and agricultural fields.
Neither of these features, nor the degree of anthropogenic disturbance, were
included in our models, yet may be good predictors of the local distribution of
RIFA. Given the extent of the invasion of RIFA into climates in the US unlike
those that it inhabits in South America, it seems unlikely that this mechanism
alone accounts for the large difference between the Forward- and Reverse-
ENM predictions. If microhabitat selection was responsible for the differences
in predictions, it seems plausible that RIFA should occur in disturbed habitats
in colder and drier regions in its native range.