Biofuel feedstocks are being selected, bred, and engineered from nonnative taxa to have few resident pests, to tolerate poor growing conditions, and to produce highly competitive monospecific stands—traits that typify much of our invasive flora. We used a weed risk-assessment protocol, which categorizes the risk of becoming invasive on the basis of biogeography, history, biology, and ecology, to qualify the potential invasiveness of three leading biofuel candidate crops—switchgrass, giant reed, and miscanthus (a sterile hybrid)—under various assumptions. Switchgrass was found to have a high invasive potential in California, unless sterility is introduced; giant reed has a high invasive potential in Florida, where large plantations are proposed; miscanthus poses little threat of escape in the United States. Each biofuel crop shares many characteristics with established invasive weeds with a similar life history. We propose genotype-specific preintroduction screening for a target region, which consists of risk analysis, climate-matching modeling, and ecological studies of fitness responses to various environmental scenarios. This screening procedure will provide reasonable assurance that economically beneficial biofuel crops will pose a minimal risk of damaging native and managed environs.
Growing energy demands, a desire to reduce reliance on fossil fuels, and greater awareness of climate change have led both state and federal governments to pursue alternative energy sources. Biomass-derived energy has been pursued for decades in the United States and Europe, but recent renewed public and political interest has sparked explosive growth in the biofuel industry. The United States initiated a research program in the late 1970s to identify candidate crops for dedicated biofuel production, whereas Europe began biofuel research in the 1960s (Lewandowski et al. 2003). However, a recent surge in bio-based fuel research has incited concern regarding rapid adoption of novel crops that may become invasive pests (Raghu et al. 2006). Herbaceous and woody species are being selected, bred, and transformed for desirable agronomic traits, including tolerance to drought, salinity, and low-fertility soils, as well as increased aboveground (harvestable) biomass and enhanced competitive ability to reduce fertilizer, irrigation, and pesticide use. However, the very traits that characterize an ideal biofuel crop also typify much of our invasive flora. Indeed, the most promising biofuel crops are nonnative to the regions proposing cultivation, compounding the potential risk of future invasions. For example, California and the Pacific Northwest are pursuing switchgrass (Panicum virgatum L.), which is native to most of North America east of the Rocky Mountains; a private firm in Florida is initiating a biofuel program centered on the Eurasian giant reed (Arundo donax L.), so-called e-grass; and Europe and the United States are screening Asian miscanthus hybrids (Miscanthus × giganteus) (Lewandowski et al. 2003).
Many invasive species have horticultural or agronomic origins with long periods of cultivation that precede their escape, naturalization, spread, and subsequent environmental impacts (Mack 2000). A classic example is kudzu (Pueraria montana [Lour.] Merr. var. lobata [Willd.] Maesen and S. Almeida), first promoted by the federal government as a forage species and later widely planted for erosion control (Forseth and Innis 2004). The rooting structure, perennial habit, and extraordinary growth rate of kudzu made for an ideal erosion mitigator, although these same traits fostered its eventual escape and dominance in the southeastern United States. The Southeast met a similar fate with johnsongrass (Sorghum halepense [L.] Pers.), introduced as a forage crop and now a noxious weed in 19 states. The sequence of selection and breeding for horticultural, agronomic, or soil stabilization purposes, cultivation in a naïve environment, followed by escape and subsequent environmental or economic calamity, often describes many of our most invasive species (Reichard and White 2001). Therefore, the quandary is how to balance the economic benefits of growing nonnative crops for bio-based energy while minimizing the risk of cultivating the next noxious weed or invasive plant.
In an effort to curtail the introduction of future invasive species, many researchers are developing preintroduction screening protocols that sort nonnative species on the basis of risk-assessment criteria. These risk-assessment tools are science-based protocols designed to identify likely invaders and benign species, and reject or accept them for introduction, after consideration of the taxon's biology and ecology, climatic requirements, history, and biogeography relative to the target region (Pheloung et al. 1999). The most widely adopted weed risk assessment (WRA) protocol was designed for Australia and New Zealand (Pheloung et al. 1999); its derivatives have been tested also in Hawaii (Daehler and Carino 2000), Florida (Gordon et al. 2006), the Czech Republic (Kŕivánek and Pyšek 2006); and a subsequent variation was tested in Australia (Caley and Kuhnert 2006). Each protocol was developed and validated using existing data sets of known invasive and benign species, with their respective accuracy ranging from 96 to 100 percent for “major” invasives (i.e., rejecting invasive plants), and 79 to 100 percent for noninvaders (i.e., accepting noninvasive plants). However, the accuracy of these protocols is much lower when considering introduced species of presumably minor impact. Nevertheless, screening nonnative species through a science-based risk-assessment protocol before importation produces a net bioeconomic gain (Keller et al. 2007), and should become standard protocol as a first step in preintroduction evaluations of nonnative species.
In response to the economic and environmental incentives for low-input biofuel crops and the desire to prevent future invasions, we screened the leading candidates for biofuel feedstock crops in the United States—giant reed, switch-grass, and miscanthus (a sterile hybrid)—using a validated WRA protocol to qualify their risk of invasion under various assumptions (e.g., the role of domestication) and scenarios (e.g., sterile cultivars). Our aim was to address the following three questions: (1) Would proposed biofuel feedstock species pass current standards for entry into nonnative regions (i.e., earn an “accept” rating)? (2) Could potential invasibility be reduced (or enhanced) through genetic modification? and (3) How do proposed biofuel feedstock species compare with other nonnative invasive or weedy species of similar life form and habitat preferences?
Weed risk assessment
Giant reed, switchgrass, and miscanthus were screened through the original WRA protocol (Pheloung et al. 1999), which we modified slightly for each target region. The WRA contains 49 questions in a macro-driven spreadsheet, with each question receiving points ranging from −3 to 4 (see www.agric.wa.gov.au for details). A final score is obtained by adding individual scores. Thresholds for “reject,” “accept,” and “evaluate further'” are set on the basis of minimizing the number of false positives (i.e., rejecting a benign species) and false negatives (i.e., accepting an invasive species) while maximizing accuracy. We left the categorical thresholds unchanged (> 6, reject; < 1, accept; 1–6, evaluate further), as is standard practice. An “evaluate further” result suggests that more information is needed before the nonnative species can be allowed entry into the target region. We answered questions in relation to the target area (e.g., switchgrass for California), and we used the criteria for answering questions sensuPheloung and colleagues (1999), namely “standard WRA.” All questions were answered using only information available in the primary literature. Questions 2.01 and 2.02 of the WRA concern the climatic suitability of the target region relative to the environmental tolerance of the taxon. Pheloung and colleagues (1999) suggest answering these questions using a climate-matching model (e.g., CLIMEX) to assess target-region suitability, but if this analysis cannot be performed, they recommend assuming a high match. This assumption is most robust in regions with high environmental variation (e.g., Australia), and less so in regions of limited climatic variability (e.g., southern Florida). We were unable to perform climate-matching modeling on the species examined here, so we made the assumption of high climatic suitability unless we had evidence to the contrary. When clear answers to the remaining questions were not available, we left the questions unanswered. Therefore, the results were a conservative risk estimate, because answering “I don't know” does not influence the final score.
Switchgrass is a rhizomatous perennial C4 grass with a native range spanning most of North America east of the Rocky Mountains, with local adaptation resulting in two distinct ecotypes (lowland and upland) that vary phenotypically with latitude of origin (Parrish and Fike 2005). Switchgrass was included in the initial screen for biofuel crops in the United States in the 1970s, and was determined to be the “model bioenergy species” by the Department of Energy (McLaughlin and Walsh 1998), largely a result of the broad adaptability and genetic variation available (Sanderson et al. 2006). Three decades of breeding have generated dozens of cultivars and varieties, many of which produce dense stands, tolerate infertile soils, and readily regenerate from vegetative fragments (Parrish and Fike 2005). Switchgrass yield trials are currently being conducted throughout its native range, but also in the introduced ranges of the Pacific Northwest and California.
Using the standard WRA, we were able to answer 38 of the 49 questions for switchgrass. California conditions earned an “evaluate further” response (table 1). This was surprising, because there are no records of switchgrass escaping cultivation in its introduced regions of Europe, Australia, and the Pacific Northwest of the United States (Parrish and Fike 2005), despite the description of several biotypes as “very aggressive” (Fransen et al. 2006). However, a lack of established escapes might be an artifact of the relatively short length of time in cultivation, as they may still be in a lag phase (Kowarik 1995).
The first question in the WRA asks whether the species is domesticated. A “yes” response favors acceptance (or reduced risk of invasion), under the assumption that domestication generally reduces the inherent weediness of wild types (Pheloung et al. 1999). Although this may be generally true, especially for agronomic and horticultural species, we argue that in the case of biofuel crops, the trend is actually the opposite: the direction of selection (breeding) is for enhanced “weedy” characters (Raghu et al. 2006). For example, agronomic crops (e.g., corn) have been bred to produce high yields, contingent on high inputs of nutrients and water in a cultivated environment, which consequently reduces competitive ability in a natural setting. This artificial shift toward overproduction of grains or fruits (as compared with the wild type) has indeed reduced the potential weediness of domesticated taxa because of their dependence on cultivation, thereby reducing the chances of establishing outside the agronomic environment. In contrast, an ideal biofuel crop should require little human subsidization of water, nutrients, or pesticides, allowing cultivation on marginal lands.
Feedstock cultivars are being bred to thrive in conditions that mimic natural environments, not the artificially rich agronomic conditions experienced by traditional crops. We assert that this selection trajectory in biofuel crops will increase the probability of survival without cultivation and the ability to invade natural environments. Additionally, by answering “no” to question 1.01, we are not biasing the assessment toward increased invasiveness, because “no” yields 0 points while “yes” yields −3 points. We reassessed switchgrass using the WRA with only question 1.01 (“Is the species highly domesticated?”), changed the answer from “yes” to “no,” and answered all other questions the same as “standard WRA.” This single difference changed the outcome from “evaluate further” to “reject” (table 1). Therefore, because the assumption of reduced weediness in domesticated biofuels may be unsuitable when screening species introduced for biomass production purposes, this question was answered “no” in subsequent analyses.
Switchgrass possesses many characteristics that favor escape from cultivation in the nonnative range of California—namely, high seed production, the ability to regenerate from vegetative fragments, rapid growth rates, and broad environmental tolerance (Parrish and Fike 2005). These autecological characters, combined with even minor environmental subsidization (i.e., fertilization and irrigation), possible contamination of planting and harvesting equipment, and unchecked seed rain during transportation from field to energy-conversion facilities, result in a high probability of invasion (see box 1). To further investigate the role of seed propagule pressure in potential switchgrass invasiveness, we screened switchgrass under the assumption that seed production was inhibited as a result of genetic modification (table 1). The “sterile genotype” alternative changed the WRA result from “reject” to “accept,” suggesting that most of the invasion risk in switchgrass is related to potential seed escape. However, as Pheloung and colleagues (1999) pointed out in their initial protocol, genotypes that are modified, whether through genetic modification or hybridization, must possess stable traits with no chance of reversion. Despite the high risk of invasion, many switchgrass cultivars require more precipitation than occurs in most of California (Fransen et al. 2006). Much of California receives less than the documented 64-centimeter annual requirement of current cultivars, with most precipitation occurring in the dormant winter months. This should greatly reduce the risk of escape, though climate-matching modeling (e.g., Holt and Boose 2000) and ecological analyses in the target region should be employed to confirm this conclusion.
Giant reed is a rhizomatous perennial C3 grass native to East Asia and the Mediterranean coast, and is considered one of the largest grass species (Lewandowski et al. 2003). Its rapid growth rate, tolerance to disturbed sites and infertile soils, and growth form have fostered widespread introduction for uses such as musical instrument construction, textiles, ornamentals, building materials, and bank stabilization (Perdue 1958). However, giant reed has escaped and naturalized in numerous introduced regions, and is a state-listed noxious weed in California and Texas (figure 1). Despite the recognition of giant reed as an aggressive pest in many introduced regions, and the difficulty in managing established infestations (Bell 1997), a private firm in Florida is proposing an “e-grass farm” for biofuel feedstock production (Fox 2007). Although giant reed was introduced to Florida more than 100 years ago and has naturalized in 23 counties, it is not a listed invasive species according to either the Florida Exotic Plant Pest Council (www.fleppc.org/list/list.htm) or the Institute of Food and Agricultural Science (Fox et al. 2005).
Giant reed was included in the species list used to create the original WRA (Pheloung et al. 1999). It received a score of 4, “evaluate further” (table 1), but few questions were answered, and some of those were answered incorrectly. A reanalysis of the original Australian assessment was conducted for Florida in response to the proposed e-grass farm, and the plant received a “reject” response (Fox 2007). We reexamined both analyses and determined that several of the questions were answered incorrectly in both (table 1). We found adequate information to answer 37 of the 49 questions for introduction to Florida, which also resulted in a rejection response.
Unlike switchgrass, giant reed does not produce fertile seeds in North America (Johnson et al. 2006), nor does it possess much genetic variation (Khudamrongsawat et al. 2004). Even if questions about seed production are answered positively (i.e., assuming introductions occur from the native range where seed production likely occurs), the outcome is still “reject” (table 1).
The “reject” outcome comes from the naturalization of giant reed in many regions outside its native range; it is also very difficult to control once it is established, and vegetative fragments can be transported along watercourses, particularly during flooding events. Weediness or invasiveness elsewhere in the world is often cited as being the most robust predictor of invasive potential (Caley and Kuhnert 2006, Reichard and Hamilton 1997), and the WRA has weighted those questions accordingly (questions 2.04–2.05 and 3.01–3.05). The combination of widespread distribution of giant reed propagules and inherent weedy characters greatly increases the likelihood of escape and subsequent environmental damage.
Miscanthus is a genus of rhizomatous perennial C4 grasses native to the tropics and subtropics. Some species have been widely introduced for both ornamental and agronomic purposes (Lewandowski et al. 2000). Miscanthus × giganteus is a naturally occurring sterile hybrid between Miscanthus sinensis (Anderss.) and Miscanthus sacchariflorus (Maxim.) French; it has been studied for biofuel production in Europe for nearly 30 years (Lewandowski et al. 2000). Despite the minimal genetic variation for crop improvement, M. × giganteus is capable of tolerating a variety of climatic conditions. The lack of seed set considerably increases the costs of establishing the crop, as it must be clonally propagated. However, sterility also dramatically reduces the probability of escape into natural environs (Heaton et al. 2004). It is important to note, however, that one of its parental lines, Miscanthus sinensis, has a prolific history of naturalization and environmental degradation (DiTomaso and Healy 2007).
The WRA results for miscanthus in the United States yielded an “accept” rating. The relatively minor risk of invasion from miscanthus is attributed primarily to the lack of seed production (Lewandowski et al. 2000). Additionally, after nearly three decades of field research across Europe, there has not been a single report of escape beyond cultivation (Lewandowski et al. 2000).
Agronomic and invasive characters
In addition to the primary concern of harvestable yield, agronomic crops are selected for traits conferring enhanced resource-use efficiency, broad pest tolerance, and tolerance to stressful growing environments (e.g., drought). Historically, selection trajectories for more robust crops have not resulted in the creation of novel weeds, despite the fact that most crop species are not native to the region cultivated (Diamond 1997). All major agronomic crops (corn, soybean, wheat, rice) require environmental subsidization through cultivation (i.e., water, nutrients), without which they would not survive, thereby mitigating weediness. This could be a consequence of the primary selection target—harvestable yield, which is typically fruit or grain. Selecting for enhanced yield has caused a concomitant loss in competitive ability and overall robustness, thereby requiring pesticide use and nutrient addition. However, in the case of dedicated biofuel crops (i.e., cellulose-based energy), the marketable yield is the entire aboveground biomass. To be economically competitive, biomass energy crops need to be perennial and highly efficient (regarding water and nutrients), and they need to establish rapidly into highly competitive monocultures, reallocate resources to belowground perennating structures during senescence, and harbor few pests or diseases (Raghu et al. 2006). Consequently, the leading candidates for bio-based energy are nonnative rhizomatous perennial grasses (Lewandowski et al. 2003), which require minimal human intervention for survival and establishment of productive stands.
To complement the WRA results, which suggest that certain traits may enhance invasion risk, we qualitatively compared the agronomically important traits that might contribute to invasiveness for three leading biofuel crops against those of established invasive or agricultural weedy species that are phylogenetically related or have a similar life form and habit (table 2). Johnsongrass (Sorghum halepense L.), like switch-grass, was first cultivated as forage, but subsequently escaped and has become one of the costliest weeds in the world, including in the southern United States (Warwick and Black 1983). Johnsongrass and switchgrass share nearly every characteristic (table 2), which is sobering, considering that johnsongrass is a listed noxious weed in 19 US states. Common reed (Phragmites australis [Cav.] Trin. ex Steud.) is native to the wetter areas of North America and Eurasia, but an introduced European genotype is currently displacing the native genotypes in North America (Saltonstall 2002). Giant and common reed share invaded habitats in their introduced ranges, and they both display extraordinary growth rates and biomass production that threaten the health of riparian corridors and wetlands. Like switchgrass and johnsongrass, giant and common reed also share most agronomic or invasive characteristics (table 2). Miscanthus sinensis is currently acknowledged as an invasive species in North America and Europe because of its rapid growth rate, broad environmental tolerance, and prolific production of wind-dispersed seeds (Tateno 1995). This comparison of proposed biofuel feedstock crops and analogous invaders should inspire a cautious approach in our march toward a bio-based energy supply—especially one founded in nonnative species. Our admittedly academic exercise demonstrates the frightening congruence between desirable traits from the agronomists' perspective and those of entrenched invasive plants.
There have been numerous attempts to qualify “invasiveness” (e.g., Pyšek et al. 1995, Williamson and Fitter 1996). Despite these attempts, no master character list has emerged, most likely a consequence of the interaction between the introduced species, the habitat to which a species is introduced, and the propagule pressure on that habitat (Barney and Whitlow forthcoming). However, within a given environment, there are characteristics that would confer an advantage to an introduced species relative to the resident community—for example, lack of herbivores or diseases, broad environmental tolerance, or efficient resource use. The WRA protocol attempts to qualify the impact certain plant characteristics would have relative to the target region (Pheloung et al. 1999). According to our analysis, sterility alone is enough to reduce the invasion risk of switchgrass and miscanthus to an acceptable level, while having no effect in giant reed (table 1). However, we would be remiss to conclude that introducing sterility to biofuel crops would reduce the invasive potential to a negligible level, as numerous invasive species lacking sexual reproduction are serious invaders (e.g., Myriophyllum aquaticm [Vell. Conc.] Verdc., Oxalis pescaprae L., and Polygonum cuspidatum Sieb. and Zucc. in the United Kingdom), including giant reed, the proposed biofuel crop assessed here. This complexity necessitates a genotype-specific analysis of potential invasiveness in each target region that combines risk assessment with ecological analyses.
Balancing economic gains and environmental consequences
The potential economic benefits are too great to prevent widespread introduction of nonnative species for biofuel purposes (Hill et al. 2006). However, nonnative species should not be introduced indiscriminately. Rather, candidate crops should be introduced in an environmentally responsible manner that maximizes ecosystem suitability, is economically advantageous, and minimizes invasion risk. Risk assessment, as demonstrated here, should not be the last word on feed-stock invasibility, and crops with an “accept” rating should still be screened through subsequent ecological analyses. We recommend a system of preintroduction screening for each proposed biofuel feedstock, regardless of WRA results, in specific target regions, a system that combines risk assessment, climate-matching modeling, cross-hybridization potential with related taxa, and in situ ecological analyses with agronomic trials. This parallel assessment should be performed for all candidate genotypes and cultivars, as well as for transformed genotypes of native species, because ecological interactions can vary widely within a species (e.g., Casler et al. 2004). Additionally, preintroduction assessments should be performed for each unique target region, as genotype-by- environment interactions are unpredictable and ubiquitous (Scheiner 1993). Acceptability in the target region must be viewed in the mosaic of environments in which a species can occur, as high-risk habitats may be invaded by means of long-distance dispersal events (e.g., through transportation vehicles). In addition, a management plan that demonstrates complete and unequivocal eradication of each feedstock should be in place before commercialization. This pre-introduction protocol, combined with producer scouting of field edges and adjacent habitat for escapes, will allow for early detection and rapid response. A circumspect approach to expanding our energy portfolio with nonnative feedstocks will benefit both society and our natural resources.
We wish to thank Alison Fox and three anonymous reviewers for providing valuable feedback on this manuscript.
The probability of a nonnative species establishing into a habitat is proportional to the number of propagules (seeds or other reproductive structures) released into that system (see Lockwood et al.  for a cogent review of propagule pressure and invasion success). Establishment probability will vary with the species, the habitat, and the flux of propagules (see the figure). Propagules can be introduced either intentionally (for cultivation, e.g.), or accidentally (e.g., falling off a truck). Biofuel crops will be cultivated on massive spatial scales with the intentional introduction of seeds or other propagules for stand establishment. Unintentional propagule release is possible during harvesting, bailing, and transporting to conversion facilities, as feedstocks are typically harvested after natural senescence when mature seeds remain attached. Ecosystems directly surrounding feedstock fields are most susceptible to propagule introductions, while roadsides and adjacent systems are also susceptible if feedstocks are transported without cover (e.g., flatbed trucks).