There is general recognition of the importance of propagule supply in invasion dynamics, including aspects of both organism quantity and quality (Williamson 1996
; Ruiz et al. 2000
). When controlling for source and recipient regions, previous studies suggest that large inocula are more likely to lead to establishment than smaller ones (Simberloff 1986
; Robinson & Edgemon 1988
; Grevstad 1999
; Kolar & Lodge 2001
), and that the risk of establishment increases with inoculation frequency (Drake 1991
; Grevstad 1999
). In addition to such quantitative aspects of supply, variation in the number of invasions among geographical regions may also result from differences in source regions, which affect species composition as well as other qualities of the donor biota that arrive to different sites (Vermeij 1996
; Miller et al. 2002
For coastal ecosystems, the number of ship arrivals to a region indicates the potential for species transfers, not just via ballast water (as discussed here), but also on the hulls of vessels. Recent models have attempted to evaluate invasion dynamics and consequences of management strategies by using ship arrivals as a proxy for supply, assuming that all ship arrivals present the same degree of risk (e.g. Drake & Lodge 2004
). We caution against this approach, which ignores variation, for ships or any other vector. Instead, effective models of vectors and their associated invasion risks require an explicit evaluation and treatment of how propagule supply varies in space and time. For example, propagule supply to an individual port is a complex function by ship types and source regions of the number of ship arrivals and their associated discharge patterns.
Our analyses illustrate that not all ships are created equal with respect to propagule supply. Frequency of ship arrivals alone does not adequately characterize the ballast-mediated species transfers, as it ignores multiple key components that influence the quantity and quality of propagules. More specifically, we expect invasion risk to vary among ports as a function of ballast discharge and source, as well as broader regional biogeographic patterns of the species pool, and these attributes are not reflected by number of ship arrivals.
The operational features of ships and commerce clearly have substantial impacts upon the quantity of ballast water discharged, and thus influence invasion risk. Among ship types, clear differences exist in the overall frequency and volumes of ballast water discharge. For example, container ships dominated the overall number of foreign vessel arrivals to US ports; however, this vessel type discharges relatively infrequently and in small volumes. Thus, ports showing high numbers of arrivals do not necessarily represent ports with the highest invasion risk. Moreover, the same ship type can behave very differently in different ports. Oil tankers and bulk carriers often carry cargo in one direction and ballast water in another, causing some ports to receive large quantities of ballast discharge (20
000 metric tons) per capita and other ports to receive very little from the same ships (Smith et al. 1999
). Because of this variation, the number of ship arrivals is usually a poor proxy for assessing ballast water discharge among ports.
Importantly, ballast water communities differ strongly among ships with respect to their species composition, abundance and physiological condition. Such disparity in ballast composition results from differences in the biota present at different source ports or donor regions, as well as likely differences with season and voyage conditions. In general, the ballast community at the end of a voyage represents only a subset of the original assemblage, and how well organisms survive during ballast water transfer is determined by both donor region characteristics and specific voyage characteristics. In numerical terms, this study and other data (Lavoie et al. 1999
; Smith et al. 1999
; Wonham et al. 2001
) indicate that there are generally declines in zooplankton density during the majority of voyages, regardless of their duration or the source location. However, the magnitude of decline varies between source locations and with voyage route, and much variation still remains unexplained. The extent to which variation in survivorship results from differences in biota, transport conditions, time, or the interactions of these variables, is not clear. For example, entrained species from different donor regions exhibit different dose-response curves upon inoculation to a single site, resulting from inter-regional (quality) differences in (i) environmental tolerance or requirements of the organisms, (ii) the physiological condition of organisms upon arrival, resulting from regional differences in tolerance or conditions experienced during transport and/or (iii) the strength of biotic interactions, such as competition or predation, at the recipient region.
Environmental factors independent of propagule pressure, such as availability of suitable habitat, patch size and local environmental conditions, are also critical in determining the likelihood of a species becoming established in a new environment (e.g. Lonsdale 1999
). Thus, habitat heterogeneity, community complexity, species–habitat interactions and the stochastic nature of environmental processes probably play a significant role in determining which invasions are successful and which are not (Elton 1958
; Crawley 1986
; Simberloff 1989
; Mack et al. 2000
). Moreover, some environments simply offer poor opportunity for colonization such as those with high species richness (Tilman 1997
; Stachowicz et al. 1999
) and for many freshwater species transferred to high salinity habitats or vice versa (Smith et al. 1999
Given the complexities and global scale of invasion pathways, models can be an especially powerful approach to develop predictions and explore effects of propagule supply and management on invasion dynamics. Nonetheless, the value of such models will depend upon the extent to which they incorporate key variables, including sources of variation known to affect invasion outcomes. At the very least, our data underscore a need to explicitly evaluate several aspects of vector operation to adequately characterize propagule supply by ships. We believe this approach is broadly applicable to any vector to understand the dynamics of species transfers and to advance predictive capacity about invasion outcomes.