Agriculture Reference
In-Depth Information
comprised of two components: 1) true commission error (overprediction) and
2) apparent commission error (correctly predicted areas not verifiable as such).
If true commission error caused the overprediction along the eastern boundary
of the native range, it suggests that factors that control the eastern range
boundary (e.g., competition, landcover) were not included in the model. If the
error instead is apparent commission error, this suggests that this region is hab-
itable by RIFA, but that RIFA either has yet to be documented there, or that
RIFA is unable to disperse to this region (see below).
In contrast, the Forward-ENM model greatly underpredicted the actual
invaded range of RIFA in the US (Fig. 2B). According to this model, Florida
is under the greatest threat of invasion by RIFA because it is predicted to be
suitable habitat by a majority of the best models (as portrayed by the dark gray
regions in Fig. 2B). A minority of the best models (i.e., the light gray regions
in Fig. 2B) predicted a narrow band of potentially suitable habitat along the
gulf coast of the US (where RIFA was first introduced) and regions of the
Pacific Northwest. The Forward-ENM projections suggest that the environ-
mental data layers used to model the native range of RIFA, though able to
replicate the native distribution of RIFA fairly well, are poor predictors of the
actual invaded range of RIFA in the US.
The niche model developed using Reverse-ENM overlapped the invaded
distribution of RIFA in the US, but suffered from commission errors, espe-
cially in the Pacific Northwest and north of the distribution boundary in the
southeastern US (Fig. 2C). Other research, including physiological-based
models [22, 35] and surveys of the timeline of the spread and establishment of
RIFA [20, 21], suggests that these commission errors are likely to represent
both apparent and true commission error. Northern California and southern
Oregon are considered regions where RIFA is likely to invade [22, 35], and
thus may represent apparent commission error. Regions where establishment
of RIFA is unlikely because of winter kill of colonies caused by minimum win-
ter temperatures (e.g., parts of Tennessee, Kentucky, Virginia and the Snake
River valley of Idaho) or insufficient precipitation (e.g., the desert Southwest)
may represent model overprediction (true commission error).
The projection of the Reverse-ENM model onto South America suggests
that RIFA occupies sites in the US that are colder and drier than habitats cur-
rently occupied in its native range; as such, the projection of the native range
of RIFA was shifted southward into Argentina. Also, this model predicted that
RIFA would not occur in the northern portion of its current native range due to
a lack of similar environmental conditions in the US. In terms of niches, these
results suggest that the niche RIFA occupies in its invaded range may be dif-
ferent from the niche it occupies in its native range.
Our study is the first published attempt to model the distribution of RIFA
using ENM. Two other published studies [22, 35] have attempted to model the
distribution of RIFA at scales of analysis similar to the study reported here.
However, each used a mechanistic, physiological model of colony growth,
based on the invaded range of RIFA, to predict the potential expansion of RIFA
Search WWH ::




Custom Search