1. Early detection of biological invasions can reduce the costs of control and increase its efficacy. Although much research focuses on the appearance or establishment of new invaders, few studies target the detection of spread from established populations. Managers of natural areas have limited resources; therefore, there is need for efficient methods of quantifying the spread of likely invaders in local and regional areas.
2. We employed homeowner surveys, seedling outplanting, directed seedling searches and randomly located plots to determine whether two introduced species of Oleaceae, Ligustrum lucidum and Olea europaea, demonstrate invasive levels of recruitment in California’s Sacramento Valley. These methods are examples of low-cost approaches to examining the regional spread of non-native woody species with differing habitat requirements.
3. Homeowner surveys indicated abundant recruitment of L. lucidum in irrigated areas, with no evident decline by distance from horticultural source trees. Ligustrum lucidum seedlings established readily when planted immediately adjacent to streams, but were unable to survive summer drought when located further from the water.
4. Recruitment of O. europaea at distances >100 m from source trees was uncommon. Spread of O. europaea is rare relative to the number of reproductive individuals that have been planted in the study area; where it occurs, seedling recruitment appears largely a function of propagule pressure.
5.Synthesis and applications. Low-cost and rapid methods are essential for successful long-term monitoring of spread from populations of introduced, woody plant species. We employed high-efficiency methods of spread detection for two species of Oleaceae with invasive potential and existing populations in the study region. We detected no barriers to spread by L. lucidum in areas with elevated soil moisture and consider the species a likely riparian invader. By comparison, O. europaea shows little tendency to spread. We suggest that managers combine low-input methods and direct surveys towards habitats of conservation concern and routes of likely seed dispersal.
Woody, long-lived, exotic species can be particularly challenging for managers because the long-term dynamics of spread are often unclear initially. Such species may be present at low numbers in a given region for some time prior to active spread and invasion (Crooks & Soulé 1999; Aikio, Duncan & Hulme 2010). Factors such as climate change, post-introduction adaptation and development of novel mutualisms may influence this lag phase length and facilitate invasion long after initial introduction (Richardson et al. 2000; Hellmann et al. 2008; Whitney & Gabler 2008). There is a need for methods that can provide managers with frequent, rapid and efficient assessment of the spread of those non-native species known to be present in a given region because management resources, including time and money, are limited (Aslan et al. 2009). Under most circumstances, early detection of and rapid response to invasive species can reduce their impact, enable substantial savings in control costs and improve the likelihood of successful eradication (Wittenberg & Cock 2001; Rejmánek & Pitcairn 2002; Mehta et al. 2007; Brooks & Klinger 2009; Panetta et al. 2011). Practical approaches to invasive species detection and evaluation should therefore be quick, low cost, efficient and tailored to the target species’ characteristics (Mehta et al. 2007).
Successful plant invasion requires recipient habitat suitability (Steinmaus 2011) and effective dispersal (Murray & Phillips 2010; Rejmánek & Richardson 1996). A history of invasiveness elsewhere in the world is the most reliable known predictor of plant invasiveness (Kolar & Lodge 2001). Therefore, to evaluate the likely colonization/spread of an exotic species, we suggest that managers target local habitat types resembling areas in which the species is known to be invasive. Within such habitats, likely dispersal paths based on known vectors should be examined. Targeting survey efforts in this way will reduce the time and money required, making techniques repeatable. To illustrate this approach, we examined seedling recruitment in California’s Sacramento Valley of two species of Oleaceae: European olive Olea europaea L. and glossy privet Ligustrum lucidum W. T. Ait. Both species are frequently planted at the regional urban/rural interface and have established dispersal mutualisms with regional birds (Aslan & Rejmánek 2010; Aslan 2011). Furthermore, both are problematic invasives in habitats similar to those found in the study area. The extensive cultivated source populations of these species in the Sacramento Valley create a high potential for natural area invasion.
We directed our surveys towards at-risk habitat types and likely dispersal routes, employing a combination of methods to assess ongoing seedling recruitment. Assessment of the riparian and irrigated-land specialist L. lucidum comprised (i) a survey of homeowners as a proxy for randomly located plots in an irrigated area and (ii) planted seedlings in a riparian area. Assessment of the upland specialist O. europaea comprised (i) directed seedling surveys along fencelines and hedgerows, (ii) homeowner surveys in an irrigated source area and (iii) randomly located plot surveys in a non-irrigated, rural area. Methods were selected to be feasible for managers with limited resources and to be easily repeatable because monitoring of spread from existing populations will need to be repeated at regular intervals (i.e. every few years for long-lived species) as spread dynamics are fluid. Where managers are attempting adaptive management (Holling 1978), repeat monitoring should be used to adjust interim management as needed. We assessed (i) the strengths and weaknesses of these efficient methods of spread detection and asked (ii) how these and similar methods could enable managers to evaluate early spread of target plant species and (iii) whether either species is likely to spread into natural areas where they could become problematic.
Olea europaea produces fruits that are generally large and dark purple to black in colour. Fruit loads can be in the hundreds of thousands per tree (Aslan 2011). Olea europaea is invasive in Australia, establishing monocultures that exclude other plant species (Spennemann & Allen 2000). In the Sacramento Valley, O. europaea is frequently used as a hedgerow species among agricultural fields. It is also common as an orchard crop and a landscaping species.
The Sacramento Valley is a lowland section of north-central California (roughly 40 000 km2). Higher elevations contain chaparral and oak savannas, while grasslands and agricultural fields dominate elsewhere. The climate is mediterranean, with wet winters and dry summers. Upland species are drought adapted, but the region also hosts remnant riparian zones. These moist habitats support native species of conservation concern, including at least 18 special status vertebrate species (Hunter et al. 1999). As only an estimated 3·3% of original lowland riparian habitats remain (Hunter et al. 1999), invasion of riparian communities by species such as L. lucidum has the potential to threaten native populations that are already seriously reduced.
Riparian corridors in the Sacramento Valley pass through urban areas, linking urban and natural sites with narrow wooded strips. Landscaping species introduced to urban or agricultural areas are therefore in close proximity to remaining natural habitats. The escape of non-native species from such plantings, particularly through bird-mediated dispersal, is an ongoing concern for regional natural area managers.
Evaluating Spread: Ligustrum lucidum
We targeted irrigated and riparian habitats for evaluation of the spread of L. lucidum. Reproductive trees are largely restricted to deliberate plantings in landscaped areas. Such urban zones, composed of diverse private and public parcels and subject to intensive disturbances, are difficult to survey for seedling recruitment using traditional sampling methods. We therefore conducted a homeowner survey in Village Homes, a discrete neighbourhood in Davis, CA. Village Homes is c. 35 years old, covers about 24 ha and was initially founded to offer affordable housing for university faculty. Residences are interspersed with communal gardens. Because of its history and layout, Village Homes attracts residents with interest in gardening and research. These factors probably elevated the level of cooperation with our survey, as well as contributing to high response reliability.
In April 2009, we canvassed the full neighbourhood to map all reproductive L. lucidum individuals and randomly selected half (151) of the neighbourhood properties. We visited each selected parcel, carrying sample L. lucidum seedlings. We asked residents to estimate the frequency of L. lucidum emergence on their properties. Estimates were categorical: zero, fewer than one per year, 1–3 per year, 4–6 per year and >6 per year. We also asked residents to report how often they garden and their level of familiarity with plants.
We used a Garmin GPS unit to record the locations of surveyed parcels. Homeowners reporting that they rarely garden or have little skill in plant identification were excluded from the data set. Our final sample size of surveyed homeowners (≈ random plots) was 98. We calculated the average number of reported L. lucidum seedlings per year per parcel. To examine spread patterns of L. lucidum, we calculated the potential seed rain index (PSRI) for each plot as the sum of the reciprocals of the square roots of distances from that quadrat to each adult L. lucidum individual in the neighbourhood (after Rejmánek, Richardson & Pyšek 2005). We regressed the reported seedling frequency on the PSRI to determine whether proximity to source trees was a significant determinant of seedling occurrence. A significant regression result would indicate that seedling recruitment drops rapidly with the distance from source trees and that spread rate is limited accordingly. A non-significant regression result, on the other hand, would indicate that seeds disperse readily at least as far as the greatest distances from source trees in the neighbourhood (c. 700 m) without displaying marked dispersal decline. We collected homeowner survey data during the last two hours of the afternoon for 10 days.
Because the normal Californian summer drought is intense, we hypothesized that L. lucidum seedlings would be unable to survive the first few months of growth in natural areas more than two vertical metres above perennial water. We germinated L. lucidum seeds from local source trees in a greenhouse and planted 2-week-old individuals into Putah Creek Riparian Reserve, located <1 km from Davis, CA. We established six replicate transects, parallel to one another and perpendicular to the perennial stream, beginning at the stream edge and ending at the transition between riparian woody vegetation and upland grassland (a vertical distance of c. 4·5 m). We placed five quadrats at even intervals along each transect (vertical distances above the stream: 0, 1·14, 2·28, 3·42 and 4·56 m). Within each quadrat, we planted eight seedlings, giving a total of 48 seedlings per elevation and 240 seedlings overall. Transects were spaced equally over a kilometre-long stretch of the stream. To ensure that seedling plantings along all transects were at the same five elevations above the stream, we restricted transect placement to locations where the elevation gain across the transect was steady. This methodology was also used to evaluate survival of Triadica sebifera seedlings in the same region (Bower, Aslan & Rejmánek 2009).
Ligustrum lucidum seedlings germinated in the greenhouse on 16–18 April 2008, corresponding with the conclusion of the regional rainy season and the observed emergence of spontaneous L. lucidum seedlings in irrigated sites in Davis, California. During transplanting of seedlings on 2 May 2008, the plots were still visibly moist from precipitation. We watered outplanted individuals weekly for 2 weeks to ensure establishment and then stopped watering to evaluate the effect of the dry season on seedling survival. We visited plots weekly for the first month and biweekly for the remainder of the dry season. At each visit, we tracked seedling height, herbivory and survival. During this experiment, an unusually dense population of weedy species (Melilotus albus, Glycyrrhiza lepidota and annual grasses) grew at the water’s edge at four of the six transects. These weeds crowded or completely shaded low-elevation seedlings; therefore, we estimated weedy cover at each visit using a 20-pin frame.
The experiment terminated with the return of the winter rains on 3 October 2008 (139 days after our final watering). We assumed that all seedlings appearing healthy at this point were unlikely to succumb to drought and represented probable successful establishment. We then removed all experimental seedlings. Just before the rain, we collected soil samples from each quadrat and calculated percentage soil moisture at the end of the drought period as 100 × [(initial soil weight − dried soil weight)/initial soil weight]. We used a logistic regression to evaluate the effect of percentage soil moisture on seedling survival.
Evaluating Spread: Olea europaea
As an upland, drought-adapted species, O. europaea’s source populations include both urban plants and abundant agricultural and rural sources. Furthermore, the large size of olive fruits makes vertebrate-mediated dispersal obligatory. We searched both urban and rural source populations and took into consideration likely bird movement paths.
First, we targeted olives planted in rural/agricultural areas of Yolo County, California, where seedling recruitment and spread are not obviously occurring. We searched for seedlings along likely bird dispersal paths. We located 12 fences or hedgerows running perpendicular to reproductive olive stands and terminating at those stands. These fences and hedgerows provided continuous perch sites for birds and served as extant transects. All such transects were at least 50 m in length, but because they were existing landscape features, their individual lengths varied greatly. Eight transects were at least 75 m long, six were at least 100 m long, five were at least 200 m long, two were at least 350 m long, and the longest transect was 585 m long. Two searchers walked each transect, one on each side of the fence or hedgerow. We used a GPS unit to record the distance from the source stand to all olive seedlings. Transects were artificially truncated at the existing ends of the fences and hedgerows; therefore, statistical options for analysis of these data were limited. However, we were able to examine the qualitative pattern of spread.
We used the same procedure as for L. lucidum to conduct homeowner surveys in the Village Homes district of Davis. We gathered residents’ reports of O. europaea seedling occurrence, which generated 98 useable parcel reports (≈ plots). We used an ordered multinomial logistic regression to analyse the relationship between distance from source trees (log10-transformed) and seedling occurrence (categorized into three classes: never, in some years and every year).
One of the few areas in California in which O. europaea shows visible spread is in Upper Bidwell Park in Chico, Butte County, CA (Cal-IPC 2006). Feral reproductive olives are now found up-canyon from a source landscaping population, evidently the result of bird-mediated dispersal (C. Aslan personal observation). As this area shows recruitment, we used traditional randomly located plots to evaluate spread. We mapped all feral reproductive olive trees with a Garmin GPS unit and recorded the coordinates of the leading edge of the source landscaping stand. We established 120 randomly located, circular plots across an area defined at one edge by the source population and extending for one kilometre up the canyon from that population (well beyond the known extent of the feral population). Plots were 5 m in radius. Habitat type (chaparral, riparian, grassland or oak woodland) and number of olive seedlings in each plot were recorded. We calculated the plot PSRI as the sum of the reciprocal square roots of distances between that plot and all reproductive olive trees within 100 m of the plot. A preliminary analysis indicated that the olive seedlings were highly clumped, so we used a generalized linear model with a negative binomial error structure and log link to analyse the relationship between the number of olive seedlings (response variable), PSRI, habitat type (as a categorical variable) and perpendicular distance (log-transformed) to the creek’s edge. We started with a model that included the additive effects of PSRI, distance and habitat and then used an analysis of deviance test to evaluate whether any reduced models had greater support than the full model. This model selection approach allowed us to determine whether olive seedling densities were predicted by proximity to source trees (i.e. largely a function of propagule pressure) or by microenvironmental variables such as moisture and habitat type.
We recorded the resources (time and money) invested in each method. The linear and logistic regressions and the anova were conducted with JMP 5·0 (SAS Institute), and the multinomial logistic regression with an ordered proportional-odds logit model was implemented from the MASS package in R (R Development Core Team; http://www.R-project.org).
The average annual frequency of L. lucidum seedlings reported per residential parcel in Village Homes was 3·52, for an estimated annual recruitment of c. 1063 seedlings from a total of 12 adult L. lucidum individuals in the neighbourhood. The regression of seedling frequency on PSRI was non-significant (P =0·65; R2 = 0·002).
Among transplanted L. lucidum individuals in the riparian reserve, 71% of seedlings planted immediately adjacent to the perennial stream survived the full drought period and were in good health when precipitation returned in the autumn (Fig. 1). At higher elevations, all seedlings had died by day 115. Of those lowest elevation seedlings that did not survive, 17% were killed by visible herbivory and 13% by drought. At its peak, surrounding weedy cover averaged 93·3% (±0·03 SE) at the lowest elevation; 21·7% (±0·08 SE) at elevation 2; 6·7% (±0·04 SE) at elevation 3; 10·8% (±0·09 SE) at elevation 4; and 45·0% (±0·14 SE) at elevation 5. The average height of surviving seedlings at the experiment’s termination was 13·5 cm. The logistic regression of survival as a function of soil moisture was positive and highly significant (P <0·0001; R2 = 0·74).
We detected 445 O. europaea seedlings on 12 transects along fencelines and hedgerows. Seedling occurrence on most transects peaked a short distance away from the source stand and then decreased rapidly with increasing distance from the source stand (Fig. 2). In total, 397 seedlings (nearly 90%) occurred within 100 m of source stands. A handful of seedlings occurred at substantially greater distances: nine seedlings were found more than 200 m and three more than 300 m from any source reproductive individuals. The greatest recorded distance between a source and seedling was 363 m.
Village Homes residents reported 0·36 O. europaea seedlings per year per parcel, yielding a total annual, neighbourhood-wide recruitment estimate of 109 seedlings from 8 within-neighbourhood source trees. The regression of seedling occurrence on PSRI was significant, but explained only a small amount of the variance (P <0·0012; R2 = 0·10). The multinomial logistic regression showed clear differences in the relationship between distance from source trees and the probability of belonging to one of three classes for the number of years in which germination occurred. The probability of belonging to the ‘Never’ class tended to increase with distance from the source tree, while the probability of belonging to the ‘Some’ or ‘All’ classes decreased with distance (Fig. 3). Although the intercepts between the ‘Some’ and ‘All’ classes were significantly different (Table 1), the shapes of their curves were similar (Fig. 3). Again, seedling emergence further than 100 m from source trees was a rare event (Fig. 3).
Table 1. Ordered multinomial regression statistics for the relationship between the distance from Olea europaea seedlings to the nearest source tree and the number of years for which homeowners in Village Homes reported seedlings. Distance was log-transformed, and occurrence by year classes was: none (did not occur); some (occurred in some years) and all (new seedlings in every year). All t-values were significant (P <0·05)
In Bidwell Park, under natural habitat conditions and visible O. europaea spread, we detected 319 feral reproductive individuals. Random plots yielded a total of 451 seedlings. The effect of PSRI on seedling frequency was significant in Bidwell Park, together with distance from water and habitat (Table 2). The habitat type with the highest frequency of seedlings was riparian (13·0 ± 1·9 SE), followed by oak woodland (3·5 ± 0·25 SE) and chaparral (3·0 ± 0·48 SE). No seedlings occurred in grassland. The percentage of deviance explained by the best-supported model was 55·8%.
Table 2. Results of a negative binomial regression evaluating the effects of distance from source trees, habitat type and distance to water on Olea europaea recruitment in its spread zone in Bidwell Park, CA. Potential seed rain index (PSRI) is the sum of the reciprocals of the square roots of distances from each quadrat to all source trees within 100 m radius of that quadrat. N = 120 plots. Habitat types included chaparral, grassland, oak woodland and riparian. Parameter estimates for grassland, oak woodland and riparian were relative to chaparral. Distance to water was log-transformed. The best model consisted of PSRI, distance to water and habitat. The analysis of deviance shows the reduction in deviance with sequential removal of individual effects from the base model
Distance to water (log)
Analysis of deviance
Distance to water (log)
None of these methods required a large financial input (Table 3). The survey of random plots in Bidwell Park took the most time primarily because it is physically demanding to locate all reproductive individuals in a rugged environment (Table 3).
Table 3. Monetary and time inputs required for each detection method
Riparian seedling survivorship experiment
$75 (greenhouse rental, supply purchase)
Olea spread zone seedling survey (random plots and mapping)
Olea bird dispersal transect surveys
Our results indicate that L. lucidum is likely to spread in moist Central Valley urban and rural areas. Recruitment from urban source trees was consistently high over distances of at least 700 m in irrigated habitats, probably because fruits are small and seeds are transported by a wide variety of bird species (Aslan 2011). Seedlings adjacent to a perennial stream in a natural riparian area survived the long summer drought in spite of intense competition with herbaceous weeds for shade and light. Our results supported our expectation that the moisture requirements of the species restricted it to areas close to perennial water; however, within that low-elevation riparian zone, the study detected no barriers to spread by L. lucidum.
The methods we utilized to evaluate L. lucidum spread were very low cost and took little time. In combination, they enabled us to assess spread of the species in the source tree habitat as well as in zones where establishment is likely in natural areas. As our concern lies in dispersal of the species between these two habitats, examination of both was essential. The total time input for the two study components combined was c. 6 person-days. As the methods detected active spread, it is unlikely that they will need to be repeated in future; instead, regional managers must determine appropriate courses of action for preventing establishment of new L. lucidum populations within and along riparian areas.
At the same time, these methods carried several weaknesses. Neither method was replicated across time or space. The results are specific to a single urban area and to a single riparian system. As many land managers will be concerned with a specific natural area or riparian system, it is likely that their use of methods such as these would be confined to a limited geographic area, as was ours. However, replication would be necessary to make generalizable predictions about the behaviour of the species under a range of circumstances. Such replication would of course increase the investment required. Additionally, because L. lucidum spreads so readily over large distances, these methods did not define an outer limit to high-frequency spread from source plants; clearly, a homeowner survey over a larger area would be required to develop a dispersal curve with a distinct drop-off. However, L. lucidum is planted at such high densities in our study area that we think it would be difficult or impossible to find enough homeowner parcels located >700 m from any planted reproductive L. lucidum individual to conduct such a study. Finally, the reliability of surveyed homeowners was uncertain as their ability to identify target plants was self-assessed. Therefore, our conclusions must be interpreted with caution. At the same time, the neighbourhood we selected for our homeowner surveys contains a populace with a higher-than-average interest in gardening, which probably made our survey easier than would be the case in a neighbourhood with different characteristics. Nevertheless, we feel that both of these methods are examples of low cost, efficient approaches that land managers can utilize when seeking to repeatedly evaluate the local spread of exotic species. Even when homeowners have little familiarity with plants or the target weeds are inconspicuous, two elements of our surveys can help to make surveys successful. First, the self-assessment by surveyees of their abilities enables researchers to exclude some households from the results; if a large number of parcels must be excluded, then the overall survey area can be expanded to raise the power of the analysis. Secondly, the use of sample seedlings, which we carried during surveys, helped homeowners to identify the species: they could be directly compared with garden weeds if surveys are conducted during the emergence season.
Establishment of O. europaea at distances >100 m from source trees, in both irrigated and non-irrigated upland areas, was uncommon. This general trend held across all detection methods. High unexplained variance in the urban environment (Village Homes) may stem from the wide range of irrigation and disturbance conditions under which each property is maintained. In the rural environment (Bidwell Park), the importance of both habitat and distance to water became apparent. Seedling recruitment was highest in riparian areas with highest availability of perch trees for avian dispersers. Recruitment was lowest in grassland, where bird perches are absent, and we therefore expect reduced seed rain. Site characteristics therefore predicted seedling density although recruitment was still uncommon (slightly more than one seedling per reproductive adult across the full census).
The limited spread rate reported here indicates that establishment of new O. europaea individuals is likely to be a function of propagule pressure and can still be considered a rare event relative to the number of adult olive trees in the region. Olea europaea is therefore probably low-risk for invasiveness in much of the study region. This assessment must be qualified, however, with the acknowledgement that an active localized invasion is visible in Bidwell Park.
Resource investment in the search methods employed for O. europaea was limited to person-days. For all three methods combined, a total of 18 person-days were required. The majority of this time was invested in the randomly located plots sampling for seedlings in Bidwell Park, as it was necessary to locate and map all reproductive individuals over a 1-km2 area with hilly topography and brushy vegetation.
Only a single urban area was surveyed in the homeowner survey, hence the results cannot be statistically applied outside that area. However, the other two methods used to assess O. europaea spread have sufficient replication to make broader application of the results more defensible. Directed seedling searching encompassed 12 hedgerows and fencelines from across a county-wide region. As these are long-lived species and unmanaged seedlings will remain in place for years, it is not necessary to search in different seasons. At Bidwell Park, the area of interest was the zone of invasion itself as it is visibly different from O. europaea dynamics elsewhere in the region. Therefore, the large number of quadrats examined across Bidwell Park was sufficient to draw conclusions relevant to the full feral population.
It does not appear that active management designed to prevent broad invasion by O. europaea is currently warranted. However, spread dynamics may change over time. Repeating these surveys every decade or so will be necessary to monitor the population. If future spread is detected, management responses can then be mounted before the regional population has grown too large.
We encourage land managers to take into account several considerations when examining the early spread of potentially invasive woody species. (i) Woody species invasion may be slower and more subtle than invasion by fast growing species. This time represents a window for early identification of invasion. As management resources are usually limited, highly efficient, targeted methods of spread detection are critical. However, not all low-cost survey methods are equally effective. A combination of methods, enabling results to be checked for consistency, will help managers ensure that no single method misleads them. (ii) Managers should identify habitats of conservation concern in which spread must be evaluated as a priority. Methods of evaluating spread must take all relevant habitats into account. (iii) The dispersal mechanisms of the plant in question are crucial to spread (Rejmánek 2011). The distances over which spread must be evaluated depend on whether seeds are dispersed by wind, birds, mammals, gravity or water (Schupp 2011). After establishing the method(s) of dispersal, directed searching should focus on dispersal pathways. (iv) If sampling is not replicated, the general relevance of the results will be limited, and the conclusions can be applied only to the regions examined. Managers wishing to apply their results more broadly should replicate their methods in space and time.
Although the invasive species assessed in this study are in the same family and are bird dispersed, they have very different habitat requirements and spread dynamics. Our results give rise to the following conclusions that may guide management of natural areas in this system: first, L. lucidum spreads easily in this system and establishes readily in moist sites, so is high risk for invasiveness. It should not be planted near natural areas. Where young trees are established in riparian areas, they should be removed before they become foci for new invasions. By contrast, O. europaea has been in the area for a long time, but its seedlings are still largely restricted to locations close to reproductive individuals. Substantial spread and invasiveness are unlikely under current conditions. Even if spread does occur, it is likely to progress slowly and with few seedlings relative to the number of reproductive adults. Therefore, it is a low priority for direct management intervention.
We thank A. Aslan, A. Bennett, M. Bower, R. Epanchin-Niell, S. Krause, M. Hufford, C. Liang and S. Veloz for assistance with data collection. L. Bolick, E. Chow and R. McKee helped with seedling transplanting. Thanks to T. Ellsworth Bowers for his mapping genius. We are grateful to E. Rejmankova and S. Castle for equipment loan and training. The managers of Putah Creek Riparian Reserve and Bidwell Park gave permission to conduct this research on their properties. C.E.A. was supported by a National Science Foundation Predoctoral Fellowship. Funding for this work was provided by a Montana State Center for Invasive Plant Management Seed Money grant and a research grant from the UC Davis Biological Invasions Integrative Graduate Education and Research Traineeship Program.