Introduction

In 1981, a large marsh plant belonging to the genus Echinodorus (Alismataceae) was collected from Pensacola, Florida. This specimen was later noticed to represent a species previously unknown in the USA and a new name Echinodorus floridanus Haynes & Burkhalter (Florida burhead) was established (Haynes & Burkhalter, 1998). No other locations of occurrence are known to the USA, and the only existing population consisting of several hundreds of individuals is bisected by a highway (Haynes & Burkhalter, 1998; Haynes & Hellquist, 2000). Hence, the species was considered to be a rare endemic and a status for threatened or endangered species was suggested by Haynes & Burkhalter (1998), and the species was mentioned to be of a conservation concern in Flora of North America (Haynes & Hellquist, 2000).

However, this status now appears to be an artifact of erroneous taxonomy (Lehtonen, 2008). Recent advances in Echinodorus systematics (Lehtonen, 2006; Lehtonen & Myllys, 2008) revealed several unnatural species and subspecies level rankings in previous classifications, which resulted in a partial revision of the genus (Lehtonen, 2008). The close affinity of Echinodorus floridanus with a South American species E. grandiflorus was already suspected by Haynes & Burkhalter (1998), and the conspecific status of the two was confirmed by morphological (Lehtonen, 2006) and molecular evidence (Lehtonen & Myllys, 2008). This ledto the formal synonymization of the name E. floridanus under E. grandiflorus (Cham. & Schltdl.) Micheli (Lehtonen, 2008).

Echinodorus is one of the most popular ornamental plants in aquaria (Kasselmann, 2001) with a cultivation history dating back into the early twentieth century (Rataj, 1978). After a short boom of production in South America, a large-scale production began in Southeast Asia, Europe, and USA (Lehtonen & Rodríguez Arévalo, 2005). Due to the favorable climatic conditions, Florida became the center of aquatic plant nurseries in the North America (McLane, 1969).

In the USA, numerous naturalized aquatic plant species became introduced in the nature via the aquarium pathway (Les & Mehrhoff, 1999; Kay & Hoyle, 2001; Maki & Galatowitsch, 2004; Cohen et al., 2007; Keller & Lodge, 2007), and many are causing immense threats to the native species and ecosystem services (Les & Mehrhoff, 1999). These species are of various geographical origins, and represent widely divergent taxonomic groups (Les & Mehrhoff, 1999). Since it has been recognized that the family Alismataceae is over-represented by invasive species (Daehler, 1998) and Florida provides exceptionally suitable landscapes for various nonindigenous aquatic plants (McLane, 1969), it is possible that E. grandiflorus possesses a threat of becoming a nuisance in Florida.

Further risk is introgression of nonindigenous genes into native populations, commonly recognized as one of the main threats for biodiversity (Seehausen et al., 2008). Even quite remotely related Echinodorus species are commonly crossed in order to produce new cultivars for the ornamental use in aquaria (Kasselmann, 2001). Two native Echinodorus species are present in Florida: E. cordifolius (L.) Griseb. and E. berteroi (Spreng.) Fassett (Haynes & Hellquist, 2000). In this article, the origin of the E. grandiflorus population in Florida, its invasive potential, and possible risks of hybridization with the native species are discussed.

Materials and methods

Ecological niche modeling has become a common practice to predict species’ potential ranges and their invasive potential in nonindigenous areas (Peterson & Vieglais, 2001). This author used Maxent ver. 3.3.0 (Maximum entropy modeling; Phillips et al., 2006; software available at http://www.cs.princeton.edu/~schapire/maxent) to predict the potential range of E. grandiflorus in southern USA. In order to produce an ecological niche model for E. grandiflorus, the author used 46 georeferenced occurrence points from the species native range, based on the herbarium specimens collected between 1925 and 2004 and deposited at BA, BM, BR, H, K, MVFA, MVFQ, MVJB, MVM, SI, TUR, UC, and UNA [herbarium codes according to Thiers (continuously updated)]. South America was used as a background to build the model, which was then projected onto southern USA to predict the potential distribution.

The niche models were based on 19 bioclimatic variables (Table 1) from high resolution (2.5 arc minutes) climatic surfaces (Hijmans et al., 2005; http://www.worldclim.org). The climatic data are based on records of the 1950–2000 period (Hijmans et al., 2005). For evaluating model, performance data were divided into training data (75% of occurrence points) and test data for model validation (25% of occurrence points). Since the prediction may depend upon the data partitioning, this author replicated the modeling 10 times using repeated subsampling. Linear, quadratic, and hinge functions of the predictor variables were included in the models. Convergence threshold was set to 10−5 and maximum iterations at 500. The results were outputted as logistic values, which give an estimate between 0 and 1 of probability of presence.

Table 1 Bioclimatic variables used in Maxent modelling
Full size table

Other environmental variables, such as topography and availability of habitats, could have been incorporated into the model as well. Within its native range, E. grandiflorus is the most common along rivers, river deltas, and other lowland areas (pers. obs.), which also typify Florida and southern USA more broadly. Aquatic habitats possibly suitable for the species are also common in the area, as can be observed from the wide occurrence of closely related E. cordifolius and more distantly related E. berteroi in the southern states (Haynes & Hellquist, 2000; Lehtonen, 2008). Since idea of the study is to broadly limit an area where the species potentially could sustain itself, the more detailed habitat-level information is not as relevant for this study, and consequently is not included in the model.

In addition to previously mentioned collections, herbarium specimens collected from Floridian aquarium plant nurseries were found from herbaria AAU and FLAS. These specimens were not included in the niche model, but they were used to date the cultivation history of E. grandiflorus in Florida.

Results

The oldest herbarium specimen representing a wild population of E. grandiflorus in Florida was collected in 1981 (Wilhelm 9464 USF!), and further collections from the same population have been made in 1986 (Burkhalter 10310 FSU), 1989 (Burkhalter & de Graaf 11539 FSU), 1997 (Haynes & Burkhalter 9717 UNA!), 2000 (Reese 1 UNA!), and 2002 (Junge s.n. M!). The population is located at sea level in a Taxodium-dominated swamp forest (Haynes & Burkhalter, 1998), thus resembling the native gallery forests habitats. However, these are not the only specimens of E. grandiflorus collected from Florida. At least one other herbarium collection of the species has been made from the aquarium plant nursery: McClade s.n. 19/07/1985 (AAU!) from Broward County, with Brazil mentioned as the original source. Ridings & Zettler (1973) studied Aphanomyces disease in the Floridian aquarium plant nurseries and its effect on the different cultivated Echinodorus species. Their voucher collections are located in FLAS (accession numbers 109518–109524), but does not include E. grandiflorus. Hence, it appears reasonable to believe that E. grandiflorus was not yet imported into Florida at the time of the Aphanomyces invasion in the early 1970s.

Maxent modeling yielded an average test AUC (area under the curve, see Phillips et al., 2006) of 0.959 for the replicated runs, with a standard deviation of 0.061. The prediction maps averaged over all the 10 models are presented as logistic outputs in Figs. 1 (projected into native range) and 2 (projected into southern USA). The logistic output gives an relative probability of presence (Phillips & Dudík, 2008). The models did not predict Florida to be generally suitable for E. grandiflorus, but a relatively narrow coastal region from Houston to New Orleans appeared as more suitable. Three variables with the largest contribution to the Maxent model were mean temperature of the wettest quarter, precipitation seasonality, and precipitation of driest month (Table 2). According to the jackknife test, annual mean temperature is the variable including most useful information by itself.

Fig. 1
figure 1

Native locations of E. grandiflorus used for Maxent bioclimatic modeling, with a predicted potential range in South America. The prediction is averaged logistic output of 10 Maxent models, each using randomly selected 25% of data points in model evaluation. The color scale indicates relative probability of occurrence

Full size image
Fig. 2
figure 2

Potential range for E. grandiflorus in southern USA. The prediction is averaged logistic output of 10 Maxent models, each using randomly selected 25% of data points in model evaluation. The only known wild population in Pensacola, Florida (marked with a dot), was not incorporated in the modeling. The color scale indicates relative probability of occurrence

Full size image
Table 2 The percent contribution of the 10 most important environmental variables to the Maxent model
Full size table

Discussion and conclusions

Several Echinodorus species have outlying populations remote from their main distribution area. This pattern can be found for at least the following species: E. berteroi has its main distribution in North America and the Caribbean Islands, but several isolated populations are known from Ecuador, Peru, Paraguay, and Argentina (Lehtonen, 2008); E. horizontalis has its main distribution in western Amazonia, but isolated populations have been found in the mouth of the Orinoco and Amazon rivers, in Guyana, and one collection from central Amazonia (Rataj, 1978; Lehtonen, 2008); E. longipetalus has its main distribution in Paraguay and southern Brazil (Lehtonen, 2008), with two known collections from Suriname (Rombouts 447 U!, Oldenburger et al. 292 U!), and one from Venezuela (Delascio 15335 VEN!); E. cordifolius, with its main distribution in the USA, but collected also from South Brazil (Lehtonen, 2008); and E. grandiflorus, with its main distribution along the Atlantic coast of South Brazil, Uruguay, and Argentina, but one population in Florida (Lehtonen, 2008).

Most of these outlying populations are located in uninhabited and remote areas and are therefore probably natural. Furthermore, several Echinodorus species are distributed in various Caribbean Islands, or have otherwise wide ranges in South America, suggesting well developed long-distance dispersal capacity. However, old cultivation history may have changed distribution of some species, such as E. horizontalis in Amazonia (see Lehtonen & Rodríguez Arévalo, 2005). This is even more probable for E. grandiflorus in Florida; the species is a well-known aquarium plant (Kasselmann, 2001). In their review of nonindigenous aquatic plants of New England, Les & Mehrhoff (1999) found that 88% of the species had their origin in cultivation, whereas only 12% of the species had been accidentally introduced or dispersed by themselves. Aquatic plant growers in Florida have intentionally introduced exotic species in the wild to guarantee constant supply and low production costs (McLane, 1969). Therefore, it seems most likely that E. grandiflorus in Florida originated either by escaping from cultivation, or it was intentionally planted.

The risk of E. grandiflorus invasion in Florida seems to be low. Several lines of evidence support this view: the species distribution modeling revealed no areas with high predicted suitability in Florida. However, the coastal region of Texas, Louisiana, Mississippi, and Alabama apparently could support the species. Historical evidence suggests a low invasive potential as well; despite almost 30 years of existence in Florida, E. grandiflorus still inhabits only one relatively small patch. Nevertheless, it should be taken into consideration that some aquatic plants have not become invasive until decades after the initial introduction, and a common means of invasion in a large number of species have included intentional planting and escapes from the cultivation (Les & Mehrhoff, 1999). Hence, even though E. grandiflorus has not been an aggressive invader in Florida, new populations may become repeatedly established via the aquarium pathway, and this would be a risk especially in the more suitable coastal regions of Texas, Louisiana, and Alabama.

Changing climate, or introduction of new genes, for example, through hybridization with native Echinodorus species, could trigger more rapid expansion in the future. Two Echinodorus species are native to southern USA: E. cordifolius and E. berteroi (Haynes & Hellquist, 2000). Both of these species are common in the area predicted to be the most suitable for E. grandiflorus in the USA (Haynes & Hellquist, 2000). Echinodorus grandiflorus is known to produce offspring in crossings with closely related E. longiscapus Arechav. and E. floribundus (Seub.) Seub. (Rataj, 1970). Echinodorus cordifolius belongs to the same group of closely related species (Lehtonen & Myllys, 2008), thus possibly allowing gene transfer between E. grandiflorus and native Echinodorus species in Florida. The risk naturally gets higher if more E. grandiflorus populations become established via aquarium pathway or range expansion.

The distribution predictions presented are based on bioclimatic variables only. The real potential distribution area is affected by a multitude of other factors as well, such as habitat availability, dispersal capacity, biotic interactions, and so on. These factors are much harder to model, and were not included here. It is here assumed that the presence of closely related species, obvious dispersal capacity of the genus and common outbreaks of various nonindigenous aquatic plants in southern USA are strong indicators that the landscape itself would allow the range expansion of E. grandiflorus, but only if the species can tolerate the prevailing climate. So far, E. grandiflorus is present as only one small population in Florida and can be quite easily eradicated, if considered appropriate.