Limiting the damage by non-indigenous species requires rapid determination of current and potential distributions and vectors of dispersal, and development of appropriate management measures. The emerald ash borer ( Agrilus planipennis ), a wood-boring beetle native to South-East Asia, was first reported in the Great Lakes region during summer 2002. The beetle poses an enormous threat to native ash ( Fraxinus ) species of North America, as untreated trees in infested areas of Ontario, Michigan and Ohio suffer high mortality. We demonstrate that the borer has spread in North America through a combination of diffusive range extension, associated with local flights, and by long-distance 'jump' dispersal associated with human movement of infested sapling or contaminated firewood. Probability of infestation was inversely related to distance from borer epicentres but positively related to the size of human population centres. At least 9 of 39 populations that were first reported in Michigan during 2004 cannot be accounted for by local diffusion, raising the possibility that other unidentified mechanisms may be contributing to the dispersal of the beetle. In the absence of quarantine, by 2005 all of Michigan's lower peninsula was contained within the boundaries of potential diffusive range expansion. Infested ash saplings also were introduced from Michigan to Maryland during 2003, and subsequently transplanted to five sites in Maryland and Virginia. Quarantine and eradication measures have had mixed results: in the south-central USA, the species appears on the brink of eradication, whereas its distribution has continued to spread during 2005 in the Great Lakes region despite extensive containment and quarantine measures. Quarantine success in the Great Lakes region is encumbered by multiple dispersal vectors, larger borer population sizes and by the more extensive geographical distribution that was achieved prior to implementation of control measures.

Population geneticists and community ecologists have long recognized the importance of sampling design for uncovering patterns of diversity within and among populations and in communities. Invasion ecologists increasingly have utilized phylogeographical patterns of mitochondrial or chloroplast DNA sequence variation to link introduced populations with putative source populations. However, many studies have ignored lessons from population genetics and community ecology and are vulnerable to sampling errors owing to insufficient field collections. A review of published invasion studies that utilized mitochondrial or chloroplast DNA markers reveals that insufficient sampling could strongly influence results and interpretations. Sixty per cent of studies sampled an average of less than six individuals per source population, vs. only 45% for introduced populations. Typically, far fewer introduced than source populations were surveyed, although they were sampled more intensively. Simulations based on published data forming a comprehensive mtDNA haplotype data set highlight and quantify the impact of the number of individuals surveyed per source population and number of putative source populations surveyed for accurate assignment of introduced individuals. Errors associated with sampling a low number of individuals are most acute when rare source haplotypes are dominant or fixed in the introduced population. Accuracy of assignment of introduced individuals is also directly related to the number of source populations surveyed and to the degree of genetic differentiation among them (F(ST)). Incorrect interpretations resulting from sampling errors can be avoided if sampling design is considered before field collections are made.

Abstract. Human introduction of nonindigenous species constitutes a serious threat to many ecosystems, particularly lakes. Recent attempts to predict invasions have focused on the supply of propagules of nonindigenous species to recipient ecosystems from source populations. Here we develop a spatially explicit ''gravity'' model to test this concept for Bythotrephes longimanus, a crustacean waterflea from Eurasia that is rapidly invading lakes in Ontario, Canada. The gravity model predicted spread of Bythotrephes based upon seven identified risk factors (e.g., use of contaminated fishing or boat anchor line) that may allow dispersal of either live individuals or their resting eggs from invaded to noninvaded lakes, as well as based on the spatial arrangement of invaded and noninvaded lakes in Ontario. Discriminant analysis of lake gravity scores successfully identified invasion status for 74% of 170 inland lakes. A retrospective analysis of 31 invaded lakes revealed that the order in which lakes were invaded was directly related to the magnitude of vector inflows from invaded sources. Analysis of the dominant vector inflow to each invaded lake revealed a ''stepping stone'' pattern in which at least five lakes were sequentially invaded from the source population in Lake Huron. One invaded lake (Muskoka) apparently served as an invasion ''hub,'' resulting in up to 18 additional direct and 17 indirect invasions. Species spread occurred via a combination of dominant, local diffusion (median distance 12.5 km) and rare, long-distance (Ͼ100 km) dispersal. Eleven of 131 lakes that were not invaded in 2000 were reported invaded in 2001. Gravity scores of these lakes were significantly higher than those of other noninvaded systems, indicating that susceptibility to invasion can be related to the magnitude of vector inflows. A GIS model based on gravity scores indicated that distribution of Bythotrephes is expected to expand to eastern and northwestern Ontario, although most new invasions are expected to occur in the central region of the province. Our results indicate that quantitative analysis of human dispersal vectors provides a robust starting point with which to assess vulnerability of discrete ecosystems to invasion. Management efforts focused on reducing the number and magnitude of human-mediated dispersal vectors may reduce the rate of invasion of new ecosystems.

1. Biodiversity assessments of lakes depend on the ability to identify the complement of species present, although the degree of sampling required is often uncertain. We utilise long-term data to predict rotifer species richness in three habitats in three Polish lakes using rarefaction sampling methods. 2. Richness in littoral and psammon habitats did not saturate, even with up to 130 samples. Highest richness was observed in psammon habitat (119 species) in Lake Mikolajskie, followed by littoral habitat in Lakes Łuknajno (114 species) and Kuc (110 species). Littoral habitats in Lakes Łuknajno (56%) and Kuc (51%) had the most species not shared with other habitats in the same lake. 3. Species richness (Chao2) estimates ranged between 44 for pelagic and 135 for psammon habitat in Lake Mikolajskie, to 100 for psammon and 137 for littoral habitat in Lake Kuc, and 65 for pelagic and 162 for littoral habitat in Lake Łuknajno. Whole lake estimates were 167, 205 and 171 species, respectively, for these lakes, higher than the 150 to 160 species predicted by Dumont and Segers (Hydrobiologia, 1996, 341, 125). 4. Using standardised sampling, richness was significantly higher in littoral than either pelagic or psammon habitats. Contrasts of standardised rarefaction curves revealed that richness in Lakes Kuc and Mikolajskie was described as well by littoral-only or psammononly samples, respectively, as by those randomly drawn from across all habitats in the lake. 5. Species richness estimates for Lake Mikolajskie were highest in summer, followed by autumn and spring. Interannual estimates differed by up to 427%, nearly an order of magnitude greater than maximal seasonal variation of 70%. 6. Results indicate that much higher sampling intensity is required to establish species richness than is presently carried out in most lakes. Because many species can be detected only with very intensive sampling, conservation programmes must consider sampling intensity when designing studies.