Models of plant invasion risk over-estimate impact

Background/Question/Methods

Habitat suitability models developed for non-native, invasive species often implicitly assume that projected invasion risk equates to risk of impact.  However, recent studies have shown that modeled habitat suitability for some native species does not equate well to species abundance.  For invasive plants, the distinction between establishment risk (potential range) and impact risk (potential range of abundance) is critical for risk assessment and application for management.  I compiled occurrence records for 9 problematic invasive plants in the western U.S. from the Global Biodiversity Information Facility, a commonly used source of herbarium records.  Occurrence records were also compiled regionally from state and country agencies that manage invasive plants.  I collected abundance estimates at a 4 km spatial resolution based on the expert knowledge of invasive plant managers in 311 counties across the western U.S.  I compared occurrence points to abundance estimates to test whether occurrences tend to be located in areas of high invasive species abundance, thereby making models based on occurrence more likely to project impact risk rather than establishment risk.  I also created habitat suitability models (HSMs) using the three datasets and compared the overall areas of predicted suitability.

Results/Conclusions

Sixty percent of the time, herbarium occurrences were located where the species was not detected by experts, while only 26% coincided with locations of high abundance. Regional occurrences were located in areas where the species was not detected 32% of the time, and on high abundance 42% of the time.  Herbarium records, which are used ubiquitously for invasive species habitat modeling, are strongly skewed towards areas with low abundance.  As a result, habitat suitability models are likely to do a poor job of modeling impact risk.  Indeed, habitat suitability models based on herbarium records encompassed 89% of land area suitable for abundance, but nearly doubled the area of estimated impact risk (48% false positive rate). Habitat suitability models based on regional occurrences did not overestimate risk of abundance (11% false positive rate), but encompassed only 64% of area suitable for abundance. Invasive plant managers dealing with existing invasive species are likely to be much more concerned with identifying risk of high abundance than risk of low abundance.  Currently, invasion risk models based on herbarium occurrence records vastly overestimate potential invasive plant abundance, thereby overestimating impact.