Estimating invasion dynamics with geopolitical-unit level records: performance and similarity of common methods using both simulated data and a real case

Estimating invasion dynamic is important to the management of invasive species, and geopolitical-unit level data are usually the most abundant and available records of invasive species. Here, for the first time we evaluated performances and similarities of eight common methods to estimate spread pattern and spread dynamic of invasive species with geopolitical-unit level data, and assessed impacts of variations in geopolitical-units on each method using simulated spread data. We also formulated a concave hull boundary displacement method (i.e., CEB) and an area-based regression method (i.e., AER) for estimating spread with geopolitical-unit data. Three regions with different sized counties in the United States (U.S.) were selected to conduct simulations and three spread scenarios were simulated. R2 and root mean square error were used to evaluate the abilities of all methods to estimate spread. Correlation coefficients were used to assess the similarity pattern of all methods. Finally, kudzu bug Megacopta cribraria, an invasive insect in the U.S., was used as a case study to test the generality of some results concluded from the simulated research. We found the CEB and two regression methods consistently estimated the right expansion patterns. Two boundary displacement and two area-based regression methods estimated highly correlated spread and were the best four methods, among which CEB had the best estimation. Distance-based regression methods are sensitive to irregularity and stochasticity in spread, and the minimum spread distance method had low ability to estimate spread. The case study showed consistent results with the simulated research. Both regression and boundary displacement methods can estimate spread patterns, overall rate, and spread dynamics of invasive species. Boundary displacement methods best estimate spread rates and dynamics; however, for spread without clear infestation outlines, area-based regression methods can be good alternatives.