Landscape-level persistence and distribution of alien feral crops linked to seed transport

Agriculture, Ecosystems and Environment 203 (2015) 119–126

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee

Landscape-level persistence and distribution of alien feral crops linked to seed transport Ross Meffin a, * , Richard P. Duncan b,a , Philip E. Hulme a a b

Bio-Protection Research Centre, 85084, Lincoln 7647, New Zealand Institute for Applied Ecology, University of Canberra, ACT 2601, Australia.

A R T I C L E I N F O

A B S T R A C T

Article history: Received 22 June 2014 Received in revised form 19 January 2015 Accepted 27 January 2015 Available online 14 February 2015

To assess the biotic and abiotic drivers of feral crop persistence, the occurrence and size of alien Brassica populations across an agricultural landscape in Canterbury, New Zealand, were surveyed over three years. Measures related to propagule input and site conditions were recorded and their role in explaining population occurrence and persistence assessed through GLMs and proportional-hazard models. Many Brassica populations were transient, with about 60% of populations disappearing within two years. New populations were founded at a rate that compensated for those that disappeared, and were more likely to occur along transportation routes and near seed companies, suggesting they established from seed spillage. Larger populations and those growing where habitat conditions were similar to those in which Brassica are cultivated had higher probabilities of survival. Without anthropogenic seed input to found new populations, Brassica spp. are unlikely to persist in this landscape beyond ten years. To avoid overestimating the extent of naturalised populations over time it is important to account for local population extinctions. The abundance of feral crops that occur as casuals in the landscape, along with other aliens that are maintained by external seed inputs, could be controlled by managing propagule sources. In themselves, casual populations are unlikely to facilitate gene flow or act as sources of further population spread. ã 2015 Elsevier B.V. All rights reserved.

Keywords: Anthropogenic dispersal Exotic Feral crops Propagule pressure Roadside vegetation Survival analysis Weed

1. Introduction Understanding the factors that drive variation in the temporal and spatial distribution of introduced plants is central to plant invasion biology (Pyšek and Hulme, 2005). A key question is often whether populations of alien and feral plants are self-maintaining and capable of supplying propagules for further spread, or whether individual populations are transient, and a species can persist in the landscape only through continual external seed inputs which found new populations at a rate higher than that of population extinction. The conditions allowing a population to establish can differ from those allowing it to persist (D’Antonio et al., 2001). Three main factors are likely to affect the likelihood of persistence of plant populations. The first is population size. Most newly founded populations comprise few individuals, and extinction can result from stochastic variation in environmental conditions and

* Corresponding author. Tel.: +61 8 8215 0045. E-mail addresses: rmeffi[email protected] (R. Meffin), [email protected] (R.P. Duncan), [email protected] (P.E. Hulme). http://dx.doi.org/10.1016/j.agee.2015.01.024 0167-8809/ ã 2015 Elsevier B.V. All rights reserved.

demographic rates (Lande, 1993). Furthermore, when a population falls below a threshold density Allee effects can be an additional cause of extinction (Taylor and Hastings, 2005). Larger populations are more likely to persist because they are less susceptible to environmental and demographic stochasticity and Allee effects (Duncan et al., 2014). Second, spatial heterogeneity in environmental conditions will mean that different locations will be more or less suitable for population persistence, and founding populations introduced to locations where the instantaneous rate of population growth is below zero will fail (Levine et al., 2004). Populations sited where conditions are suitable for recruitment, survival and reproduction, resulting in a positive rate of population increase, will have a higher chance of persisting (Duncan et al., 2014). Third, external seed input can rescue populations that may otherwise go extinct by increasing population size and/or increasing the rate of population growth. Sources of external seed input include anthropogenic sources (Hodkinson and Thompson, 1997) and seed dispersing from neighbouring populations (Husband and Barrett, 1996). Additionally, a persistent seed bank can act as a buffer against stochastic fluctuations in environmental conditions by allowing population recovery after unfavourable years (Philippi, 1993). In summary, large population size, suitable environmental conditions, little environmental variability,

120

R. Meffin et al. / Agriculture, Ecosystems and Environment 203 (2015) 119–126

external seed inputs and persistent seed banks are all factors that favour population persistence. Agricultural regions are among the most invaded habitats (Jauni and Hyvönen, 2010) and feral populations of many crops are widespread in these regions (Gressel, 2005; Garnier et al., 2006). The population dynamics of feral crop populations are not well understood (Garnier et al., 2006; Pivard et al., 2008a; Bagavathiannan et al., 2010), and in particular little is known about the factors that determine persistence (Pessel et al., 2001). Persistent populations of feral crops can act as reservoirs for genes and bridges for gene flow (Ellstrand et al., 1999; Bagavathiannan and Van Acker, 2008), hindering efforts to contain the traits of both genetically modified (GM) and conventional crops (Bagavathiannan and Van Acker, 2008; Bagavathiannan et al., 2012). Containment of crop traits is desirable both to prevent the escape of novel genes from GM crops (Ellstrand et al., 1999) and for maintaining the integrity of genetically distinct cultivars (Gressel, 2005). Furthermore, feral crops can become problematic invaders of natural and semi-natural habitats at higher rates than are predicted for other introduced plant species (Williamson, 1994). Brassica is a widely cultivated genus, usually grown in moist, fertile, tilled and managed fields (Heenan et al., 2004; Pivard et al., 2008a; Knispel and McLachlan, 2010). Roadsides and disturbed field margins tend to be the most heavily invaded habitats in agricultural regions (Jauni and Hyvönen, 2010), and feral Brassica populations are common in roadside habitat in many parts of the world where Brassica species are cultivated (Heenan et al., 2004; Pivard et al., 2008a; Knispel and McLachlan, 2010). Feral populations often occur in the vicinity of Brassica seed transport routes and hubs, and are thought to be founded by seed escapes from trucks, farm machinery, and adjacent fields either at the time of sowing or harvest (Crawley and Brown, 1995; Pivard et al., 2008a). Brassica are reliant on disturbance for successful recruitment, suggesting populations are unlikely to persist beyond a year or two without regular disturbance (Crawley et al., 1993; Crawley and Brown, 1995). Evidence from field studies suggests that persistent feral populations are widespread in regions where Brassica are cultivated (Squire et al., 2011) and that around half of populations may persist via a persistent seed bank and local recruitment (Pivard et al., 2008a), with populations capable of persisting for eight years or more (Pessel et al., 2001). This has led to concerns regarding the environmental risk associated with GM cultivars since feral GM populations have been found in Canada, the UK, Australia, France, Japan and the USA (Schafer et al., 2011), and these could impede the containment of gene flow. In this study, the status of feral Brassica populations on the Canterbury Plains, New Zealand, was determined to assess whether they pose a threat as persistent and potentially expanding populations. Specifically, a detailed field survey aimed to determine whether Brassica populations were: (a) persistent and selfmaintaining; (b) able to persist but only as part of a larger metapopulation; or (c) were mostly sink populations maintained by external seed input. Distinguishing among these possibilities is important for assessing alien distributions and abundances (Pergl et al., 2012), their potential for increase (Mack et al., 2000), and identifying appropriate management strategies (Pessel et al., 2001; Pyšek, 2005). Environmental covariates were also measured to identify factors related to population presence, and persistence from one year to the next. Data from an earlier survey in the same region (Heenan et al., 2004; Peltzer et al., 2008) allowed us to estimate population persistence over the last ten years. In line with previous work, it was expected that most feral Brassica populations would be associated with disturbance and propagule availability (Crawley and Brown, 1995) and that persistence would be a function of population size, suitability of the environment for plant growth, availability of seed inputs and the size of the seed bank.

2. Methods The study was conducted in rural districts of the Canterbury Plains, New Zealand. The climate is temperate, with a mean annual precipitation of 740.9 mm and mean annual temperature of 12.2  C (1981–2010, Darfield EWS, NIWA). Frosts are frequent in spring and winter, and inland areas may be covered by snow in winter. Much of the area is pasture, although Brassica crops (notably kale, B. oleracea acephala; turnip, B. rapa rapa; and swede, B. napus napobrassica) are also grown as fodder, particularly over winter. There is an emerging biofuel industry which uses oilseed rape (B. napus napus) seeds. Furthermore, there is a seed multiplication industry which produces B. rapa,B. napus, B. oleracea and B. juncea seeds for international export. The area is covered by a network of roads, ranging from minor unpaved roads to sealed state highways. These roads have a verge, ranging in width from approximately 2 to 5 m, where the vegetation reflects the species grown in neighbouring fields. Perennials dominate, particularly alien grasses (Lolium perenne and Dactylis glomerata) and clovers (Trifolium repens and T. pratense). Feral Brassica populations (mostly B. rapa, with some B. napus, B. juncea and B. oleracea) also occur, although these are comparatively rare (Heenan et al., 2004; Peltzer et al., 2008). Management of road verges ranges from frequent mowing and herbicide application (particularly around drainage ditches and fence-lines) to no management. 2.1. Field surveys From 2010 to 2012 annual surveys of roadside Brassica populations were conducted in September and October, coinciding with the peak flowering time for Brassica. To make the results comparable to surveys conducted in 2003 and 2005 of roadside Brassica populations in the same area (Heenan et al., 2004; Peltzer et al., 2008), a similar sampling regime was used. Surveys were conducted from a motor vehicle travelling at approximately 70 km h 1. Brassica produce tall inflorescences with conspicuous yellow flowers that are easily identified at this speed; no other plant species in the study area produce a similar floral display. The survey revisited each of 50 randomly located, 3  3 km squares that were previously surveyed in 2003 and 2005 (4.7% of the total study area), which involved travelling along all roads in each square and recording any Brassica populations encountered. The shortest route was then used to travel from one square to the next. Thus these routes were considered to be representative of the landscape, and any Brassica populations encountered en route between squares were recorded (Heenan et al., 2004; Peltzer et al., 2008). The same route (total length = 2200 km) was surveyed in each year from 2010 to 2012. For each Brassica population encountered, variables associated with propagule availability, population size, and environmental conditions were recorded (Table 1 and Supplementary material, Appendix A). To contrast the characteristics thought to influence the probability of population presence, in 2010 the same variables recorded at sites where Brassica were present were also recorded where Brassica were absent (10 sites in each of the 50 surveyed squares = 500 absence sites in total, Supplementary material, Appendix A, Table 1). The absence sites were selected by dividing the total road length in each survey square by 10, and then stopping at this distance interval along the roads within the survey square. No Brassica populations were recorded at the absence sites in 2011 or 2012, and the measurements taken at these sites in 2010 were used to contrast with the presence sites in all three years. To determine whether a seed bank might be important for the persistence of Brassica populations, soil samples were collected in March (autumn) 2011 from each population present in spring 2010.

R. Meffin et al. / Agriculture, Ecosystems and Environment 203 (2015) 119–126

121

Table 1 Variables recorded for each Brassica population encountered on the Canterbury Plains during field surveys, 2010–2012, and how they are hypothesised to influence the probability of population presence and/or persistence (Supplementary material, Appendix A). Density of seeds in the seed bank was only recorded in 2010. Levels of factor variables are shown in parentheses. Variable

Hypothesis tested in models

References

Road type (unsealed, sealed, External propagule input from traffic increases probability of presence and persistence. Traffic volumes, (Crawley and Brown, 1995) state highway) numbers of seed transportation trucks, and numbers of escaped seeds are greatest on state highways, least on unsealed roads. Distance to nearest seed company

External propagule input associated with proximity to seed companies increases probability of presence (Crawley and Brown, 1995) and persistence.

Brassica in adjacent field (Y/ External propagule input from seeds spilled during sowing increases probability of presence and N) persistence.

(Pivard et al., 2008a; Pivard et al., 2008b)

Distance to nearest population

Propagule input from nearby populations increases probability of presence and persistence (as in a metapopulation).

(Garnier et al., 2008; Knispel and McLachlan, 2010)

Density of seeds in soil

Higher density of below ground propagules increase probability of persistence.

(Philippi, 1993)

Population size

Larger populations less susceptible to stochastic effects, Allee effects. Increased probability of persistence.

(Lande, 1993)

Water course (Y/N)

Provides suitable and stable of conditions for Brassica. Increases probability of presence and persistence. (Truscott et

Soil depression (Y/N)

Provides refuges from mowing. Increases probability of presence and persistence.

(Wilson and Clark, 2001)

Disturbed (Y/N)

Provides bare sites for Brassica establishment and recruitment. Increases probability of presence and persistence.

(Crawley and Brown, 1995, 2004)

Mowed (Y/N)

Prevents growth and reproduction. Decreases probability of presence and persistence.

(Wilson and Clark, 2001)

Sprayed with herbicide (Y/ N)

Prevents growth and reproduction. Decreases probability of presence and persistence.

(Knispel and McLachlan, 2010)

Dominant vegetation type (annual/perennial)

Turnover in annual vegetation provides bare sites for Brassica establishment and recruitment. Increases (Vandermaarel and Sykes, probability of presence and persistence. 1993)

al., 2005)

Mean summer precipitation Higher summer precipitation increases probability of presence and persistence.

(Thuiller et al., 2005)

Mean annual temperature

(Thuiller et al., 2005)

Higher annual temperatures increase probability of presence and persistence.

Ten randomly located 100 mL soil samples were collected from a 2  2 m plot at the centre of each population. These were bulked to provide a single sample from each population (n = 121). Each bulked sample was then spread evenly to a depth of approximately 2 cm over a bed of sterile seed germination mix (60% peat, 40% sterilised pumice, plus 2 kg Osmocote exact mini (16-3.5-9.1), 4 kg Dolomite and 1 kg Hydrafloin per m3) on a tray within 48 h of collection (Gross, 1990). The trays were kept in a glasshouse and watered sufficiently to keep moist. Every two weeks any germinated seedlings were pricked out and the number of Brassica seedlings recorded, until no further germination occurred. Following current practice for Brassica, no freezing or stratification treatment was administered to the soil samples (de Jong et al., 2013; Gruber et al., 2014).

2.2. Analysis All analyses were conducted using the statistical software R (R Development Core Team, 2012). To test which factors could explain the probability of Brassica population presence in the landscape, a separate model was fitted for each of three years (2010–2012), using the function glm, with Brassica presence/absence at each site as the binary response variable, which was treated as Bernoulli distributed (McCullagh and Nelder, 1989). Explanatory variables related to propagule availability and site suitability for Brassica were included to assess which variables were related to the probability of Brassica population presence

(Supplementary material, Appendix A, Table 1). Distance to the nearest Brassica population in the previous year was omitted from the model for 2010 because no data were available on the locations of Brassica populations in 2009. A model was fitted including all explanatory variables and then this was simplified using step-wise backwards selection based on AIC values to arrive at a final model for each year. The average yearly survival probability of Brassica populations was estimated using a Cox proportional hazard model (CPHM), fitted using the function coxph in the package survival. This method was used to accommodate censored data (Cox, 1972), i.e. it was unknown whether populations recorded in 2010 were founded in that year or had survived from previous years, or whether populations recorded in 2012 would survive to 2013. First, a CPHM was fitted without any explanatory variables, providing an estimate of population survival probability and the associated uncertainty (Cox, 1972). A second CPHM was fitted to determine which covariates could explain variation in survival probability, including all of the explanatory variables used in modelling the probability of Brassica presence/absence (Supplementary material, Appendix A), but with two additions. First, larger populations were expected to have higher survival probabilities due to lower susceptibility to environmental and demographic stochasticity and Allee effects (Taylor and Hastings, 2005). To test this, the mean population size for each population over the three years was included as an explanatory variable. Second, factor variables that varied among years were included by using the mean of the observations over the three years, providing a measure of the probability of the site occurring in each state in any given year (i.e.

122

R. Meffin et al. / Agriculture, Ecosystems and Environment 203 (2015) 119–126

Table 2 Summary of the number of Brassica populations recorded on the Canterbury Plains field surveys in each of three years, 2010–2012, showing the total numbers of populations, turnover in each year, and the percentage of populations belonging to each taxon recorded.

Populations: total Total extinct in previous year Total founded in previous year Brassica rapa (%) oleifera (%) chinensis (%) rapa (%) Brassica napus (%) napus (%) napobrassica (%) Brassica rapa  napus (%) Brassica oleracea (%) Brassica juncea (%)

2010

2011

2012

mean  s.e.

121 – – 64.5 48.8 8.3 7.4 24.0 21.5 2.5 4.1 0.8 6.6

177 68 124 57.6 47.5 9.0 1.1 29.9 28.8 1.1 5.6 1.1 5.6

190 143 156 68.4 47.4 19.5 1.6 27.4 27.4 0 0.5 1.1 2.6

150  19 105  21 130  13 63.2  2.3 47.9  0.5 12.3  3.6 3.4  2.0 24.9  2.5 25.9  2.2 11.2  0.7 2.8  1.2 3.0  2.0 86.1  1.4

that a site was mowed, sprayed, disturbed, had annual/biennial vegetation or was adjacent to a Brassica crop). A model was fitted including all explanatory variables and then this was simplified by AIC backwards step-wise selection to attain the final model. To test whether the density of seeds in the soil at the end of the 2010 growing season affected the probability of population survival until the following year, a binomial GLM was fitted with the response variable being whether a population present in spring 2010 survived until spring 2011, and the number of seeds germinated per L of soil in soil samples collected in autumn 2011 as the explanatory variable.

[(Fig._1)TD$IG]

Differences between the most common species, B. rapa, and other Brassica species were assessed by fitting models to the data for B. rapa alone, and then to the data for all species combined. The CPHM including explanatory variables was also re-fitted, including species as an additional explanatory variable to test for differences among species in population survival. Few differences were found; see Supplementary material, Appendix B. 3. Results A total of 401 Brassica populations were recorded during the 2010–2012 surveys (Table 2). Populations were widespread but infrequent; there was on average around one population per 15 km of road surveyed, or 6.4 individuals per km of road. The total number of populations recorded in each year varied and tended to increase over the course of the study (mean  s.e., 150  19). Around 60% of surveyed 3  3 km squares contained no Brassica populations in any year. Most populations were small, comprising fewer than 10 individuals, with few populations comprising more than 1000 individuals, and the largest estimated to contain 5000 individuals (Fig. 1). Much of the increase in the number of populations over time comprised an increase in the number of small populations of a few individuals (Fig. 1). Few populations survived for more than one or two years following first detection. Of the populations recorded in 2010, only 44% were still present in 2011, and only 16% were present in 2012 (Table 2, see also Fig. 1 black bars vs. grey bars). There was a large turnover of populations from year to year: around half of the populations present in 2010 had disappeared by 2011, but around

Fig. 1. Frequency histograms of Brassica populations recorded on roadsides on the Canterbury Plains in three annual surveys, 2010–2012, classed by the number of Brassica individuals in the population (log scale). Black bars indicate populations that had disappeared the following year; grey bars indicate populations that survived until the following year (not shown for 2012, no survey was conducted in 2013).

R. Meffin et al. / Agriculture, Ecosystems and Environment 203 (2015) 119–126

123

[(Fig._2)TD$IG]

twice as many new populations were founded. Around three quarters of the populations present in 2011 had disappeared by 2012, but a similar number of new populations were founded (Table 2). 3.1. Influence of propagule pressure Brassica populations were significantly more likely to be found on well-used transport routes, being most frequent on state highways, followed by sealed roads and then unsealed roads (Table 3). Of all the variables tested, road type had the largest effect and this was consistent among years. Brassica populations were also more likely to be found close to seed companies (Table 3). Around 85% of populations occurred within 20 km of a seed company, with median distances of 6.2, 7.5 and 6.1 km in 2010, 2011 and 2012, respectively. The estimated probability of Brassica presence declined sharply with distance away from the nearest seed company (Fig. 2). Neither location adjacent to a cultivated Brassica field, nor distance to the nearest Brassica population in the previous year, were significantly linked to the probability of the presence of Brassica populations. 3.2. Influence of site conditions Brassica populations were more common along watercourses and in disturbed locations than other sites in all years, while management of vegetation by mowing was associated with a lower frequency of Brassica populations (Table 3). 3.3. Survival of Brassica populations Populations had a mean probability of surviving for one year of 61% (95% CI: 56–67%), and a mean probability of surviving for two years of 44% (95% CI: 35–50%, Fig. 3). Site characteristics influencing the probability of Brassica population survival differed from those associated with the probability of presence; larger populations growing on the margins of watercourses or in soil depressions were more likely to survive (Table 4). The annual probability of populations sited on the banks of watercourses disappearing was reduced by 45% compared to other populations, while growing in a depression lowered the probability of disappearing by 56%. Each individual in a population decreased the probability of disappearing by 0.3%. Surprisingly, Brassica populations at sites controlled with herbicide sprays tended to be more persistent, with a 33% reduction in the annual probability of disappearing.

Fig. 2. The relationships between the probability of Brassica population occurrence and the distance to the nearest company handling bulk quantities of agricultural seed. Lines show relationships derived from parameter estimates of GLMs fitted to data from field surveys, 2010–2012, symbols indicate survey sites where Brassica populations were present (i.e. probability present = 1) and absent (i.e. probability present = 0).

In contrast to presence, the probability that Brassica populations survived from one year to the next was not related to variables associated with propagule availability (road type, distance to nearest seed company, distance to nearest population, presence of Brassica in the adjacent field, and 2010 seed bank density). The number of seeds found in the seed bank in autumn 2011 ranged from 0 to 73 seeds L 1 of soil (2.4  0.8, mean  s.e.) and was not significantly associated with the probability of population survival from spring 2010 to spring 2011. 4. Discussion Despite a continuous presence in the landscape, which without careful observation may give the impression of population persistence, the overwhelming majority of Brassica populations on the Canterbury plains are transient and appear unlikely to persist beyond a few years without external seed inputs. Thus, these populations pose little risk as sources of spread and invasion. Despite this, there remains the possibility of such populations

Table 3 Results of GLMs of the probability of Brassica population presence on the Canterbury Plains for each year, 2010–2012, showing all variables tested, and mean regression estimates  standard errors of significant variables. Factor variables (road type, water course, mowed, disturbed, sprayed, vegetation type, adjacent field, depression) are in comparison to the reference class (no parameter estimates). A dash indicates the variable was not retained in the model for that year. Significance: *** = p < 0.001, ** = p < 0.01, * = p < 0.05. 2010 Sealed road State highway Distance from seed company Water course Mowed Disturbed Sprayed Mean summer precipitation Annual vegetation Distance to nearest population in previous year Brassica in adjacent field Mean annual temperature Soil depression

2011

3.01  0.63*** 3.83  0.69*** 1.08  0.27*** 1.06  0.32*** 1.53  0.28*** 0.93  0.27*** – – 2.13  0.34*** – – – –

2012

2.21  0.42*** 4.83  0.67*** 1.34  0.38*** 0.92  0.43* 4.40  0.41*** 1.31  0.34*** – – – – – – –

2.72  0.46*** 4.63  0.61*** 1.06  0.33** 1.26  0.38*** 3.27  0.35*** 2.04  0.31*** 1.50  0.37*** 1.20  0.31*** – – – – –

124

R. Meffin et al. / Agriculture, Ecosystems and Environment 203 (2015) 119–126

[(Fig._3)TD$IG]

4.2. Estimates of population survival

Fig. 3. Results of a Cox proportional hazard model based on three years’ survey data (2010–2012) showing estimated survival probability of Brassica populations on the Canterbury Plains and the associated 95% confidence interval (dashed lines).

acting as short-lived genetic bridges allowing introgression of genes between crops and their wild relatives, and between crop varieties grown in different fields (Hails and Morley, 2005; Knispel et al., 2008). 4.1. Propagule sources The accidental dispersal of seeds from trucks transporting seeds is likely to be the main source of new Brassica populations given the strong association of these populations with transport routes and proximity to seed companies. It seems few roadside populations are founded by seed immigration from neighbouring fields. Seeds spilled during harvest may be an additional propagule source, although most Brassica crops observed during the surveys were fodder rather than seed crops and so generally did not set seed. There was no evidence of metapopulation dynamics, whereby populations were founded or maintained by seeds dispersing from other feral populations. Consequently, limiting external seed input into roadside habitats from transport sources, for example by fitting suitable covers to seed transport trucks, would most likely reduce feral Brassica populations in this landscape.

Table 4 Results of a Cox proportional hazard model of survival of Brassica populations on the Canterbury Plains 2010–2012, showing mean parameter estimates  standard errors, coefficient exponentials, z-values and associated p-values. The coefficient exponentials are interpretable as the proportional effect of that coefficient on the probability of a population becoming locally extinct e.g. holding all other covariates constant, a population sited on a water course has on average around 55% the annual risk of extinction of populations not sited on a water course, an increase in population size of one individual reduces the yearly risk of population extinction by 0.3%. Estimate  s.e. All species Water course Soil depression Population size Sprayed with herbicide

0.601  0.207 0.828  0.296 0.003  0.001 0.263  0.101

eestimate 0.548 0.437 0.997 0.769

z

Pr(>|z|) 2.898 2.799 1.938 2.595

0.004 0.005 0.053 0.009

In the absence of new populations establishing through external seed inputs, our results suggest that feral Brassica populations would be unlikely to persist in the long term: only 15% of populations present in 2010 survived until 2012. Nevertheless, this may underestimate survival because it is likely that some populations present in 2010 had established earlier. Accounting for this with a CPHM suggested that around 40% of populations survive for at least two years following foundation. Although there is no continuous record of populations recorded in the original 2003 survey, 5% of the populations observed in 2010 occurred at or in close proximity to sites where Brassica was present in 2003 (Heenan et al., 2004; Peltzer et al., 2008). Whether these populations persisted throughout this period or went extinct and were recolonized later is not known, but this would suggest an upper limit of about 5% survival over ten years. Surveys that seek to quantify the persistence of plant populations often contain censored data and CPHMs are a valuable tool for analysing such data. The low persistence of feral Brassica is this system is in contrast to studies in Europe that report widespread persistence (Squire et al., 2011), with up to 50% of populations persisting for at least one year (Pivard et al., 2008a) and some for at least eight years (Pessel et al., 2001). Our results are more consistent with those of Crawley and Brown (1995, 2004),) where populations of B. napus along the M25 motorway persisted for a median of 1.5 years, with a small minority of sites occupied for as long as ten years. Crawley and Brown (1995) also identified seed spilled along major transport routes as the main propagule source founding new populations. In contrast, the system studied by Pivard et al. (2008a) consisted of mostly minor roads and most populations were founded by seed dispersing from neighbouring fields during harvest. Seed spills during harvest are estimated to occur at densities of several 1000 seeds m 2 (Pivard et al., 2008a), while those from trucks are estimated at around one tenth that density (Bailleul et al., 2012). Given the relationship found between population size and persistence, it seems likely that spills during harvest will establish larger, more persistent populations than those resulting from spills during transport, and are consequently of greater management concern. 4.3. Implications of low population survival As with Brassica in this study, other alien plant species may be reliant on continual propagule input to maintain their landscape distribution. Indeed, many aliens inadvertently introduced into the Czech Republic via seeds in wool imports failed to persist in the absence of this propagule source (Pyšek, 2005). Other cases may include weed impurities in agricultural seed (Conn, 2012; Lehan et al., 2013) and escaped garden plants (Meyer and Lavergne, 2004; Marco et al., 2010). Despite these examples, the issue of persistence of local populations of alien plants is rarely addressed in the literature. Understanding local population persistence is important from three standpoints. First, populations of casual aliens (sensu Richardson et al., 2000) are unlikely to be sources of further spread and, if control is a goal, reducing the propagule supply should lead to decline. Second, assessments of species’ distributions are often based on cumulative survey and herbarium records (Petrík et al., 2010), which could overestimate distribution if they fail to account for populations that subsequently go extinct. In the Czech republic Pergl et al. (2012) found that only half of the grid cells assumed to be occupied by the invasive species Heracleum mantegazzianum on the basis of cumulative presence data were actually occupied when resurveyed. Third, identifying populations of feral crops as persistent or not is important to assessing their potential role as

R. Meffin et al. / Agriculture, Ecosystems and Environment 203 (2015) 119–126

gene reservoirs and the extent to which they might enhance gene flow across the landscape (Garnier et al., 2006; Bagavathiannan et al., 2010). Monitoring of feral crop populations is required to effectively confine novel traits (Bagavathiannan and Van Acker, 2008), and this necessitates differentiating feral populations which persist from those that are transient.

125

of B. napus oilseed crops in the UK and Canada, which support large oilseed industries (Crawley and Brown, 1995; Knispel and McLachlan, 2010) involving the transportation of large quantities of harvested seed. By contrast, B. rapa is likely to be transported in smaller amounts in Canterbury where it is primarily used as a fodder crop. While B. napus is grown for oilseed in Canterbury, only a small amount of seed is produced (Hampton et al., 2012).

4.4. Factors influencing population survival Population size was a major determinant of the likelihood of population persistence. It is probable that larger populations are buffered against the negative effects of stochasticity and/or less susceptible to Allee effects (Lande, 1993). Those sites where Brassica tended to be most persistent were those with less environmental stochasticity and conditions consistently suitable for Brassica recruitment, growth and reproduction. The fields where Brassica are grown are usually moist, fertile, low-competition sites that are relatively undisturbed after pre-sowing tillage (Peltzer et al., 2008). It seems likely that drainage and irrigation ditches, where conditions are similar to such fields (Truscott et al., 2005), provided the most suitable conditions for Brassica persistence out of the available roadside sites. Sites periodically sprayed with herbicides to maintain bare patches around roadside objects (Heenan et al., 2004) appear to offer ruderal plants such as Brassica spp. a competitive advantage over other, predominantly perennial roadside vegetation, and may enhance local recruitment (Crawley and Brown, 2004) leading to increased persistence. Populations sited in soil depressions were also more persistent, either because site conditions such as moisture are more favourable in depressions, or because plants growing in depressions may better escape the impacts of mowing, which is a common disturbance in roadside habitat (Wilson and Clark, 2001). Models and field studies indicate seed immigration, local recruitment and the presence of a persistent seed bank increase feral crop population persistence (Pessel et al., 2001; Garnier et al., 2006; Pivard et al., 2008b; Bagavathiannan et al., 2010). In this system, neither seed immigration from spillages, neighbouring fields and populations, nor seed banks were linked to population persistence. Instead population size and site conditions acted as higher order filters, constraining persistence. Seeds of B. rapa, the most common Brassica species in our study system, are thought not to have primary dormancy (Adler et al., 1993; de Jong et al., 2013). Seeds of B. napus on the other hand, can have limited primary dormancy (Baskin and Baskin, 2001) such that cold stratification can increase germination rates by around five percent (de Jong et al., 2013), although this increased germination rate can be negated by increased mortality of nondormant seeds (Adler et al., 1993). Nevertheless, the possibility remains that our decision not to cold stratify seeds resulted in slight underestimates of seed bank densities for B. napus populations. Given the low rates of primary dormancy in B. napus and the prevalence of B. rapa in our study system, it is unlikely that cold stratification of soil samples would have altered our conclusion that seed banks have little influence on population persistence. 4.5. Frequency of populations Roadside Brassica populations were infrequent compared to similar studies in Europe and Canada, just 6.4 Brassica individuals km 1, most of which were B. rapa. In the UK Crawley and Brown (2004) reported around 50 B. napus plants km 1, and in Canada, Knispel and McLachlan (2010) recorded between 5 and 550 B. napus plants per km. The difference likely reflects the dominance

4.6. Population dynamics The number of Brassica populations recorded increased both between 2003 and 2005 (Heenan et al., 2004; Peltzer et al., 2008) and from 2010 to 2012. It seems unlikely that these increases represent a lag effect or invasion in progress as feral Brassica have occurred in Canterbury for decades (Peltzer et al., 2008) and seem to be reliant on external seed inputs to persist. Fewer populations were detected in 2010 than 2005, suggesting declines are also possible. One explanation for the observed changes in population frequency is variation around an equilibrium driven by interannual variation in climate (Crawley and Brown, 2004; Peltzer et al., 2008) and/or the quantity of seed transported. A further possibility is that participants in both studies became more effective at finding Brassica populations through time. Much of the increase from 2010 to 2012 was in populations of a few individuals which are relatively hard to detect, lending support to this possibility.

5. Conclusions Rather than being fully naturalised, in the sense of having a selfmaintaining wild population, as is generally thought (Heenan et al., 2004), Brassica on the Canterbury Plains may be better classified as casual. Presence in the landscape is maintained because the rate at which local populations go extinct is compensated for by new populations establishing, most likely via seed spills along major seed transport routes. Our results highlight that population size and local site conditions can be more important for population persistence than seed availability from external sources or seed banks. Human-assisted dispersal is an important vector of alien plant spread (Hodkinson and Thompson, 1997), which may result in aliens such as garden ornamentals (Marco et al., 2010), and those introduced as seed contaminants of commodities such as wool (Pyšek, 2005) and agricultural seed (Conn, 2012; Lehan et al., 2013), displaying similar population dynamics to those documented here. Failing to recognise such population dynamics may result in an overestimation of alien abundance and distribution, but also has management implications. Transient sink populations with little local recruitment are unlikely to lead to further spread or invasion, and such transient feral crop populations have reduced potential to act as genetic reservoirs or bridges (Garnier et al., 2006; Bagavathiannan et al., 2010). However, even where populations are transient, there can be gene flow events and feral populations can accrue multiple transgenes, for example acquiring multiple herbicide resistance (Knispel et al., 2008). Consequently, transient feral crop populations are of concern for those GM crops for which very low levels of gene flow or absolute containment are a requirement. If a goal is to limit or eradicate populations, control of seed inputs may be adequate for sink populations. To target more persistent populations, larger populations sited in drainage and irrigation ditches, and those founded during the harvest of seed crops could be prioritised for management, and the efficacy and timing of herbicide applications and mowing need to be carefully considered (Garnier et al., 2006).

126

R. Meffin et al. / Agriculture, Ecosystems and Environment 203 (2015) 119–126

Acknowledgements We gratefully acknowledge funding from Foundation for Research, Science and Technology (NZ), the Bio-Protection Research Centre, and the New Zealand Plant Protection Society. These funding bodies had no input in the execution or content of this study. We convey our sincere thanks to Yukari Izumo and Diana Clark for valuable field assistance, Robin Pouteau for assistance with GIS data extraction, and Peter Heenan and Duane Peltzer for making data from the 2003 survey available. Appendices A and B Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.agee.2015.01.024. References Adler, L., Wikler, K., Wyndham, F., Linder, C., Schmitt, J., 1993. Potential for persistence of genes escaped from canola: germination cues in crop, wild, and crop-wild hybrid Brassica rapa. Funct. Ecol. 736–745. Bagavathiannan, M.V., Begg, G.S., Gulden, R.H., Van Acker, R.C., 2012. Modelling the dynamics of feral alfalfa populations and its management implications. PLoS One 7 (6), e39440. Bagavathiannan, M.V., Gulden, R.H., Begg, G.S., Van Acker, R.C., 2010. The demography of feral alfalfa (Medicago sativa L.) populations occurring in roadside habitats in Southern Manitoba, Canada: implications for novel trait confinement. Environ. Sci. Pollut. Res. 17, 1448–1459. Bagavathiannan, M.V., Van Acker, R.C., 2008. Crop ferality: implications for novel trait confinement. Agric. Ecosyst. Environ. 127, 1–6. Bailleul, D., Ollier, S., Huet, S., Gardarin, A., Lecomte, J., 2012. Seed spillage from grain trailers on road verges during oilseed rape harvest: an experimental survey. Plos One 7. Baskin, C.C., Baskin, J.M., 2001. Seeds: Ecology, Biogeography, and Evolution of Dormancy and Germination, second ed. Academic Press, San Diego. Conn, J.S., 2012. Pathways of invasive plant spread to alaska: III. Contaminants in crop and grass seed. Invasive Plant Sci. Manag. 5, 270–281. Cox, D.R., 1972. Regression models and life tables. J. R. Stat. Soc. Ser. B (Method.) 34, 23. Crawley, M.J., Brown, S.L., 1995. Seed limitation and the dynamics of feral oilseed rape on the M25 motorway. Proc. R. Soc. Lond., Ser. B: Biol. Sci. 259, 49–54. Crawley, M.J., Brown, S.L., 2004. Spatially structured population dynamics in feral oilseed rape. Proc. R. Soc. Lond., Ser. B: Biol. Sci. 271, 1909–1916. Crawley, M.J., Hails, R.S., Rees, M., Kohn, D., Buxton, J., 1993. Ecology of transgenic oilseed rape in natural habitats. Nature 363, 620–623. D’Antonio, C.M., Levine, J.M., Thomsen, M., 2001. Ecosystem resistance to invasion and the role of propagule supply: a California perspective. J. Mediterr. Ecol. 2, 233–245. de Jong, T.J., Tudela Isanta, M., Hesse, E., 2013. Comparison of the crop species Brassica napus and wild B. rapa: characteristics relevant for building up a persistent seed bank in the soil. Seed Sci. Res. 23, 169–179. Duncan, R.P., Blackburn, T.M., Rossinelli, S., Bacher, S., 2014. Quantifying invasion risk: the relationship between establishment probability and founding population size. Methods Ecol. Evol. 5, 1255–1263. Ellstrand, N.C., Prentice, H.C., Hancock, J.F., 1999. Gene flow and introgression from domesticated plants into their wild relatives. Annu. Rev. Ecol. Evol. Syst. 30, 539–563. Garnier, A., Deville, A., Lecomte, J., 2006. Stochastic modelling of feral plant populations with seed immigration and road verge management. Ecol. Model. 197, 373–382. Garnier, A., Pivard, S., Lecomte, J., 2008. Measuring and modelling anthropogenic secondary seed dispersal along roadverges for feral oilseed rape. Basic Appl. Ecol. 9, 533–541. Gressel, J., 2005. Crop Ferality and Volunteerism, first ed. CRC Press, Boca Raton. Gross, K.L., 1990. A comparison of methods for estimating seed numbers in the soil. J. Ecol. 78, 1079–1093. Gruber, S., Weber, E., Claupein, W., 2014. Which soils are comfortable for oilseed rape seeds (Brassica napus) to survive? Plant Soil Environ. 60, 280–284. Hails, R.S., Morley, K., 2005. Genes invading new populations: a risk assessment perspective. Trends Ecol. Evol. 20, 245–252. Hampton, J.G., Rolston, M.P., Pyke, N.B., Green, W., 2012. Ensuring the long term viability of the New Zealand seed industry. Agron. N. Z. 42, 129–140. Heenan, P.B., FitzJohn, R.G., Dawson, M.I., 2004. Diversity of Brassica (Brassicaceae) species naturalised in Canterbury, New Zealand. N. Z. J. Bot. 42, 815–832. Hodkinson, D.J., Thompson, K., 1997. Plant dispersal: the role of man. J. Appl. Ecol. 34, 1484–1496.

Husband, B.C., Barrett, S.C.H., 1996. A metapopulation perspective in plant population biology. J. Ecol. 84, 461–469. Jauni, M., Hyvönen, T., 2010. Invasion level of alien plants in semi-natural agricultural habitats in boreal region. Agric. Ecosyst. Environ. 138, 109–115. Knispel, A.L., McLachlan, S.M., Van Acker, R.C., Friesen, L.F., 2008. Gene flow and multiple herbicide resistance in escaped canola populations. Weed Sci. 56, 72–80. Knispel, A.L., McLachlan, S.M., 2010. Landscape-scale distribution and persistence of genetically modified oilseed rape (Brassica napus) in Manitoba, Canada. Environ. Sci. Pollut. Res. 17, 13–25. Lande, R., 1993. Risks of population extinction from demographic and environmental stochasticity and random catastrophes. Am. Nat. 142, 911–927. Lehan, N.E., Murphy, J.R., Thorburn, L.P., Bradley, B.A., 2013. Accidental introductions are an important source of invasive plants in the continental united states. Am. J. Bot. 100, 1287–1293. Levine, J.M., Adler, P.B., Yelenik, S.G., 2004. A meta-analysis of biotic resistance to exotic plant invasions. Ecol. Lett. 7, 975–989. Mack, R.N., Simberloff, D., Lonsdale, W.M., Evans, H., Clout, M., Bazzaz, F.A., 2000. Biotic invasions: causes, epidemiology, global consequences, and control. Ecol. Appl. 10, 689–710. Marco, A., Lavergne, S., Dutoit, T., Bertaudiere-Montes, V., 2010. From the backyard to the backcountry: how ecological and biological traits explain the escape of garden plants into Mediterranean old fields. Biol. Invasions 12, 761–779. McCullagh, P., Nelder, J.A., 1989. Generalized Linear Models, second ed. Chapman and Hall/CRC, Boca Raton. Meyer, J.Y., Lavergne, C., 2004. Beautes fatales: Acanthaceae species as invasive alien plants on tropical Indo-Pacific Islands. Divers. Distrib. 10, 333–U334. Peltzer, D.A., Ferriss, S., FitzJohn, R.G., 2008. Predicting weed distribution at the landscape scale: using naturalized Brassica as a model system. J. Appl. Ecol. 45, 467–475. Pergl, J., Pyšek, P., Perglová, I., Jarošík, V., Procheş, S., 2012. Low persistence of a monocarpic invasive plant in historical sites biases our perception of its actual distribution. J. Biogeogr. 39, 1293–1302. Pessel, F.D., Lecomte, J., Emeriau, V., Krouti, M., Messean, A., Gouyon, P.H., 2001. Persistence of oilseed rape (Brassica napus L.) outside of cultivated fields. Theor. Appl. Genet. 102, 841–846. Petrík, P., Pergl, J., Wild, J., 2010. Recording effort biases the species richness cited in plant distribution atlases. Perspect. Plant Ecol. Evol. Sys. 12, 57–65. Philippi, T., 1993. Bet-hedging germination of desert annuals – beyond the first year. Am. Nat. 142, 474–487. Pivard, S., Adamczyk, K., Lecomte, J., Lavigne, C., Bouvier, A., Deville, A., Gouyon, P.H., Huet, S., 2008a. Where do the feral oilseed rape populations come from? A large-scale study of their possible origin in a farmland area. J. Appl. Ecol. 45, 476–485. Pivard, S., Demsar, D., Lecomte, J., Debeljak, M., Dzeroski, S., 2008b. Characterizing the presence of oilseed rape feral populations on field margins using machine learning. Ecol. Model. 212, 147–154. Pyšek, P., 2005. Survival rates in the Czech Republic of introduced plants known as wool aliens. Biol. Invasions 7, 567–576. Pyšek, P., Hulme, P.E., 2005. Spatio-temporal dynamics of plant invasions: linking pattern to process. Ecoscience 12, 302–315. R Development Core Team, 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria Available from: http://www.R-project.org. Richardson, D.M., Pyšek, P., Rejmanek, M., Barbour, M.G., Panetta, F.D., West, C.J., 2000. Naturalization and invasion of alien plants: concepts and definitions. Divers. Distrib. 6, 93–107. Schafer, M.G., Ross, A.A., Londo, J.P., Burdick, C.A., Lee, E.H., Travers, S.E., Van de Water, P.K., Sagers, C.L., 2011. The establishment of genetically engineered canola populations in the US. PLoS One 6 (10), e25736. Squire, G.R., Breckling, B., Pfeilstetter, A.D., Jorgensen, R.B., Lecomte, J., Pivard, S., Reuter, H., Young, M.W., 2011. Status of feral oilseed rape in Europe: its minor role as a GM impurity and its potential as a reservoir of transgene persistence. Environ. Sci. Pollut. Res. 18, 111–115. Taylor, C.M., Hastings, A., 2005. Allee effects in biological invasions. Ecol. Lett. 8, 895–908. Thuiller, W., Richardson, D.M., Pyšek, P., Midgley, G.F., Hughes, G.O., Rouget, M., 2005. Niche-based modelling as a tool for predicting the risk of alien plant invasions at a global scale. Global Change Biol. 11, 2234–2250. Truscott, A.M., Palmer, S.C.F., McGowan, G.M., Cape, J.N., Smart, S., 2005. Vegetation composition of roadside verges in Scotland: the effects of nitrogen deposition, disturbance and management. Environ. Pollut. 136, 109–118. Vandermaarel, E., Sykes, M.T., 1993. Small-scale plant species turnover in a limestone grassland – the carousel model and some comments on the niche concept. J. Veg. Sci. 4, 179–188. Williamson, M., 1994. Community response to transgenic plant release – predictions from British experience of invasive plants and feral crops. Mol. Ecol. 3, 75–79. Wilson, M.V., Clark, D.L., 2001. Controlling invasive Arrhenatherum elatius and promoting native prairie grasses through mowing. Appl. Veg. Sci. 4, 129–138.