Table E-l. Model and description (continued) 
GIS. Haltuch and Berkman (2000) use multiple regression analysis to predict percent cover by zebra mussels. The 
analysis incorporates (1) bathymetry, (2) sediment type, and (3) Side Scan Sonar data. The model indicates that 
zebra mussels spread across soft substrates and transform soft substrates to hard substrates. Mussels on soft 
substrate may serve as a “positive feedback’ for more mussels. The authors conclude that this GIS modeling can be 
used to predict spread of mussels onto soft substrates in other lakes and to examine species invasion dynamics and 
|| impacts across landscapes. 
GIS. Le Maitre et al. (1996) use GIS models to show how much water could be lost per year if plant invasions are 
allowed to continue uncontrolled. Using this information, they develop a catchment management system. Five 
processes modeled separately include: (1) fire occurrence, (2) spread and establishment of alien plants, (3) growth 
between fire cycles, (4) rainfall to run off ratio, and (5) effects of biomass on stream flow. The model shows that, 
over a 100-year period, invasive plant species cover increases from 2.4% to 62.4%. The authors recommend 
removing invasive plant species to ensure water availability. 
GIS. Wilcox et al. (2003) mapped Phragmites coverage over nine different years using aerial photos in Great Lakes 
I region. The authors conduct spatial analysis of total area covered each year. They analyze abundance changes 
using geometric or logarithmic growth equations. GIS maps show distribution was dynamic from 1945 to 1999 and 
increased exponentially from 1995 to 1999. The authors conclude that the rapid expansion of Phragmites is due to 
decreased water levels in the Great Lakes, increased ambient air temperatures, and possibly natural and human- 
induced disturbance. The authors predict that Phragmites will continue to expand at a high rate into the lower Great 
Lakes coastal wetlands given the plant’s level of invasiveness and predicted climate scenarios. 
Regression analysis. Ricciardi (2003) uses regression analysis on data from various zebra mussel (Dreissena 
polymorpha ) invasion sites to develop empirical models of zebra mussel impact. The analysis involves comparing 
I an invader’s impacts in the different regions and ecosystems where it has been introduced to determine if the 
invader’s impacts are predictable in varying habitats. The author’s results show that the zebra mussel impacts are 
predictable across various areas and habitats. The author recommends correlating models that relate invader 
abundance to physical environmental traits with models that relate invader impact to abundance. Linking these 
models to invasions will allow for predictions on which habitats will experience significant impacts from invasion, 
and management decisions can be made accordingly. 
Logistic regression analysis.* Collingham et al. (2000) use statistical models of presence/absence of three weed 
species at coarse and fine scales. The authors evaluated ability of model at one scale to predict distribution at larger 
scale. The results show some correspondence between environmental factors at different spatial scales. The authors 
recommend modeling species at more than one scale. This is important for managers, because weed control happens 
at a fine scale, but understanding processes on a larger scale is important for long-term management. For example, 
analyses show that climatic variables affect species’ ranges; thus, range may be affected by future environmental 
change. 
Multiple logistic regression analysis. Goodwin et al. (1999) conducted a multiple, logistic regression analyses to 
determine if the invasiveness of introduced species can be predicted based on widely available biological data. The 
analyses were conducted on 165 pairs of plant species originating from Europe, where one species in the pair has 
invaded Canada and the other species has not (110 pairs were used in the multiple regression analysis and 55 pairs 
were used to test predictive ability of the regression models). Three biological attributes were used in the analysis. 
The authors found that geographic range of a species is a successful predictor of invasiveness, while the biological 
I attributes tested are not. However, geographic range is likely correlated with biological traits. The authors conclude 
that predicting invasions on a species-by-species level will not adequately deal with the accidental introduction of 
species. 
Regression and Akaike’s information criteria. Marchetti et al. (2004) use logistic regression to determine the 
relationship between successful establishment and biological variables; multiple regression to evaluate the 
relationship of a measure of spread and the average abundance of an invasive species with biological variables; and 
Akaike’s information criteria (AIC) as an unbiased estimate of the regression model fit. The results show that 
different characteristics favor different stages of invasion (e.g., establishment, spread). The authors find that human 
preference affects invasion, and they recommended stopping the transport and release of non-native fish to prevent 
invasions. _ 
E-5 
