Image spectroscopy and stable isotopes elucidate functional dissimilarity between native and nonnative plant species in the aquatic environment


  • Maria J. Santos,

    1. Center for Spatial Technologies and Remote Sensing (CSTARS), Department of Land, Air and Water Resources, University of California, One Shields Avenue, The Barn, Davis, CA 95616, USA
    Search for more papers by this author
  • Erin L. Hestir,

    1. Center for Spatial Technologies and Remote Sensing (CSTARS), Department of Land, Air and Water Resources, University of California, One Shields Avenue, The Barn, Davis, CA 95616, USA
    Search for more papers by this author
  • Shruti Khanna,

    1. Center for Spatial Technologies and Remote Sensing (CSTARS), Department of Land, Air and Water Resources, University of California, One Shields Avenue, The Barn, Davis, CA 95616, USA
    Search for more papers by this author
  • Susan L. Ustin

    1. Center for Spatial Technologies and Remote Sensing (CSTARS), Department of Land, Air and Water Resources, University of California, One Shields Avenue, The Barn, Davis, CA 95616, USA
    Search for more papers by this author

Author for correspondence:
Maria J. Santos
Tel: +1 530 7520621


  • Nonnative species may change ecosystem functionality at the expense of native species. Here, we examine the similarity of functional traits of native and nonnative submersed aquatic plants (SAP) in an aquatic ecosystem.
  • We used field and airborne imaging spectroscopy and isotope ratios of SAP species in the Sacramento–San Joaquin Delta, California (USA) to assess species identification, chlorophyll (Chl) concentration, and differences in photosynthetic efficiency.
  • Spectral separability between species occurs primarily in the visible and near-infrared spectral regions, which is associated with morphological and physiological differences. Nonnatives had significantly higher Chl, carotene, and anthocyanin concentrations than natives and had significantly higher photochemical reflectance index (PRI) and δ13C values.
  • Results show nonnative SAPs are functionally dissimilar to native SAPs, having wider leaf blades and greater leaf area, dense and evenly distributed vertical canopies, and higher pigment concentrations. Results suggest that nonnatives also use a facultative C4-like photosynthetic pathway, allowing efficient photosynthesis in high-light and low-light environments. Differences in plant functionality indicate that nonnative SAPs have a competitive advantage over native SAPs as a result of growth form and greater light-use efficiency that promotes growth under different light conditions, traits affecting system-wide species distributions and community composition.


Rapid environmental changes may cause plant functional traits to become mismatched with current environmental conditions. Current species distributions reflect both present ecological sorting and past selective pressures, and range expansion or survival in new geographic locations are dependent upon traits that are preadapted to the new environment. These preadaptations are likely to have competitive and evolutionary advantages, such as those observed in nonnative species with unconstrained growth and that become invasive (Mack et al., 2007) and replace native species. What is uncertain is whether the presence of nonnative species retains or changes the functionality of the previous ecosystem to the disadvantage of native species. In this paper we examine to what extent the functional traits of native and nonnative submersed aquatic plants (SAPs) are similar in an aquatic ecosystem.

One system where environmental conditions markedly constrain plant community functionality is the submersed aquatic ecosystem (Sculthorpe, 1965). Plant functional traits can be measured at metabolic, physiological and morphological levels. Functional metabolic traits include photosynthetic pathways, the substrates used for photosynthesis, and the ability to respond to varying light intensities. SAPs have constraints on photosynthesis that are imposed by carbon availability and light in the water column (Dennison et al., 1993). To cope with the low-CO2 environment in the water column, most SAPs have a carbon concentration mechanism (CCM; Maberly & Madsen, 2002) that allows them to store carbon for photosynthesis, either from CO2 or HCO3 substrates (Van et al., 1976; Sand-Jensen, 1983). The main photosynthetic pathways for SAPs are C3, often coupled with CCMs (Maberly & Madsen, 2002), and many use other photosynthetic pathways, including C4, CAM (Crassulean acid metabolism), and C3–C4 intermediates (Keeley, 1999; Ueno, 2001; Bowes et al., 2002; Keeley & Rundel, 2003). Most SAPs are restricted by light partitioning in the water column, with excess light at the surface and low light in deeper water (Dennison et al., 1993). Some SAPs are known to switch from C3 photosynthesis in the light-limited environment at depth to a C4-like photosynthesis and also in the high-light environment of the surface (Ueno, 2001). This facultative C4-like metabolism includes some of the properties of the C4 metabolic pathway, such as fixation of CO2 by phosphoenolpyruvate into malic acid (Ueno, 2001), and no activation of the xanthophyll cycle (Peñuelas et al., 1993, 1997). C4-like metabolism (Keeley, 1999; Keeley & Rundel, 2003) has been demonstrated for some SAP species, including Hydrilla verticillata (Salvucci & Bowes, 1981), Myriophyllum spicatum (Van et al., 1976), Egeria densa (Casati et al., 2000), and potentially Cabomba caroliniana (Salvucci & Bowes, 1981).

Functional physiological traits include the distribution, arrangement and composition of plant biochemical compounds (nutrients, pigments, Chl, etc.), and how these building blocks are combined to overcome environmental limitations such as submergence, high light, salinity, or temperature. Functional morphological traits include plant organs (leaves, stems, roots) and traits such as leaf :  root ratios, morphology (entire and dissected leaves, fine and tap roots, etc.), and architecture (three-dimensional distributions of leaves and stems, leaf angle distribution, ratio of leaves to stem, leaf to root, etc.). At the leaf level, morphological and physiological traits of SAP leaves are often similar to shade adaptations (Mommer et al., 2005), and their typically spherical leaf angle distributions allow them to absorb diffuse light from all directions. There are three main types of leaves in submersed plants: blade-shaped leaves (strap-shaped, elongated or ribbon-like, which are associated with lentic and lotic environments); dissected leaves (deeply cut or subdivided leaves); and whorled leaves (three or more blades at each node), which are associated with lentic environments (Luther, 1947; Sculthorpe, 1965). The metabolic, biochemical, and morphological plasticity and diversity of SAPs make them particularly well adapted to varying environmental conditions, and therefore have high potential to spread into new habitats. The functionality of ecosystem processes can be affected by changes in plant community composition depending on whether nonnative species replace or change the functions that native species performed.

While remote sensing is effective at monitoring invaded plant communities over large spatial extents (Lehmann & Lachavanne, 1997; Elmore et al., 2003; Kerr & Ostrovsky, 2003; Cohen & Goward, 2004; Coppin et al., 2004; Asner & Vitousek, 2005; Bradley & Mustard, 2006), imaging spectroscopy is well established for remote detection of plant biochemistry, photosynthetic efficiency, and leaf morphology and canopy structure (Ustin & Curtiss, 1990; Peñuelas et al., 1993; Jacquemoud et al., 1994; Gamon et al., 1997; Zhang et al., 1997; Asner, 1998; Kokaly, 2001; Ollinger & Smith, 2005). Imaging spectroscopy (also called hyperspectral remote sensing) measures hundreds of contiguous narrow bands spanning the solar reflective spectrum from 400 to 2500 nm. The result is a nearly continuous spectrum measured in each pixel that provide the information needed to detect SAPs in the water column (Zhang et al., 1997; Underwood et al., 2006; Hestir et al., 2008), despite interactions with the water column itself (Holden & LeDrew, 2001; Han, 2002; Bostater et al., 2003; Hall et al., 2004), which obscures the characteristic SAP spectral patterns. When plant species are spectrally distinct (Fyfe, 2003), imaging spectroscopy is a likely method for mapping SAPs down to species level (Ustin et al., 2009). Imaging spectrometers capture spectral differences resulting from species morphological (leaf and canopy), biochemical (pigment concentration) and metabolic (photosynthesis) traits. When the analysis framework includes additional biophysical information, such as stable isotope data (Farquhar et al., 1989; Raven et al., 2002; Carvalho et al., 2009), the combination provides multiple lines of evidence to compare and contrast functionality across species.

We applied this combined approach to the SAP species that co-occur in the Sacramento–San Joaquin River Delta (henceforth referred to as the Delta) in California, USA. The Delta is one of the major gateways for nonnative aquatic species in the United States (Cohen & Carlton, 1995, 1998). In the last 30 yr the aquatic plant community has dramatically changed: the number of species has increased (Atwater et al., 1979; Cohen & Carlton, 1998; Bossard et al., 2000; Light et al., 2005; Santos et al., 2009), with the plant community composition now c. 50% nonnative plant species (Fig. 1; Santos et al., 2009). The assemblage today includes five native and four nonnative SAPs (Table 1). All nonnative species in the Delta have both the C3 pathway and a facultative C4-like metabolism (Van et al., 1976; Salvucci & Bowes, 1981; Casati et al., 2000), as does the native Elodea canadensis (Nichols & Shaw, 1986). Varied types of leaf morphologies are found in the Delta SAP assemblage. All nonnative species have wider and longer leaves than native SAP and distinct growth forms, with blade-like and dissected leaves evenly distributed along the stem length through the water column (Sculthorpe, 1965).

Figure 1.

Submersed aquatic plant species co-occurring in the Sacramento–San Joaquin River Delta. The top four species are natives and the bottom four species are nonnatives. Egeria densa and Myriophyllum spicatum are invasive. Note the difference in the leaf structure among the different species: wide leaves include Potamogeton nodosus, E. densa, and Potamogeton crispus; fine leaves include Elodea canadensis, Stuckenia pectinata, and Myriophyllum spicatum; and whorled leaves include Ceratophyllum demersum and Cabomba caroliniana.

Table 1.   Submersed aquatic plant (SAP) species occurring in the Sacramento–San Joaquin River Delta, CA, USA
FamilyScientific namesCommon namesCodeRootsLeafStatusa
  1. aRefers to its current status in the Delta and not to its invasibility potential.

HydrocharitaceaeEgeria densaBrazilian waterweedEGDEYesEntire, wide bladeNonnative
HydrocharitaceaeElodea canadensisWaterweedELCAYesEntire, narrow bladeNative
CeratophyllaceaeCeratophyllum demersumCoontailCEDENoDissected, whorlNative
PotamogetonaceaeStuckenia pectinataSago pondweedSTPEYesEntire, no bladeNative
PotamogetonaceaeStuckenia filiformisBroadleaf sago pondweedSTFIYesEntire, no bladeNative
PotamogetonaceaePotamogeton nodosusAmerican pondweedPONOYesEntire, floating bladeNative
PotamogetonaceaePotamogeton crispusCurlyleaf pondweedPOCRYesEntire, wide bladeNonnative
HaloragaceaeMyriophyllum spicatumEurasian watermilfoilMYSPYesDissected, whorlNonnative
CabombaceaeCabomba carolinianaCarolina fanwortCACAYesDissected, whorlNonnative

Materials and Methods

Study area

Located in central California (38°19′N, 121°36′W), the Delta is formed from the confluence of the Sacramento and the San Joaquin Rivers, and drains into San Francisco Bay (CA, USA). Its wetlands were reclaimed in the early 20th century for agriculture through construction of nonnatural islands, hydrologically connected through a reticulate network of earthen levees and channels. To avoid saline intrusion of tidal waters, counter-circuit pumping of freshwater occurs system-wide to force freshwater to extraction pumps. This creates a relatively stable aquatic environment for drinking water and agriculture and conditions ideal for growth of SAP (Cohen & Carlton, 1995; Jassby & Cloern, 2000; Lucas et al., 2002).

Airborne imaging spectroscopy

We used airborne HyMap imaging spectroscopy data acquired in June 2007 by HyVista, Inc. (Sydney, NSW, Australia) at a nominal spatial resolution of 3 m. HyMap is delivered as radiometrically and geometrically corrected image measurements in 126 spectral bands in the range 400–2400 nm with a bandwidth of 15 nm in the visible-near-infrared (VIS-NIR) and 15–20 nm in the shortwave-infrared (SWIR, 1500–2500 nm) domain.

Field measurements

Over 2000 SAP patches distributed over the 2100 km2 of Delta channels were visited concurrent with acquisition of imaging spectroscopy data. For each site we documented species composition, species percentage cover, and patch dimensions. SAPs frequently co-occur in the study area at scales smaller than the ground pixel size of 9 m2 (Santos et al., 2011). Therefore, to ensure that each pixel spectrum represents a single species rather than a mixture, we chose patches with > 60% area occupied by a single species, which resulted in 1151 field samples. No pure patches were found for E. canadensis and Stuckenia filiformis, and these species were excluded from subsequent analysis. Spectra of pure patches of each of the SAP species and of turbid and clear water (assessed with a Secchi disk) were extracted using STARSPAN (Rueda et al., 2005; for all bands for each of the 1151 field samples, one pixel per sample.

Handheld spectrometer measurements

Ten specimens of each SAP species were collected in the field and grown in aquatic tanks in a glasshouse for handheld spectrometer measurements (Table 1). An ASD Field Spec Pro FR (ASD Inc. Boulder, CO, USA) spectrometer with a 0.45 rad instantaneous field-of-view (IFOV; 0.01 m diameter) was used to collect 30 reflectance measurements of monospecific dense canopy mats for each SAP species at 20 cm above the plant surface in full intensity midday (11:00–13:00 h) sunlight. To correct for background and water influence in the spectral signature, we collected 30 measurements of tank water and 30 of tap water. Because the phenological stages of the submersed species were different, the Potamogeton nodosus sample was in advanced senescence when the other species were ready for measurement, and consequently, it was excluded from the glasshouse study. This species is unusual, as it acts as an emergent dominant with most of its canopy at the water surface rather than within the water column (Santos et al., 2011), making it functionally more similar to emergent aquatic species like water hyacinth (Eichhornia crassipes), and dissimilar to other SAP species.

We used the Spectral Analysis and Management System (SAMS; software to extract reflectance for all spectrometer bands, and screened the data to exclude bands at wavelengths < 420 nm and > 1200 nm because of noise. We resampled the glasshouse spectrometer data to the wavelength resolution of the airborne sensor to facilitate comparison using a standard Gaussian resampling model with 15 nm band spacing in ENVI 4.4 (ITT, Boulder, CO, USA).

Mapping SAPs, species differences and separability analysis

We used hierarchical Boolean classification schemes (decision trees), utilizing several spectral analysis methods (Giardino & Zilioli, 2001; Holden & LeDrew, 2001; Han, 2002) at different nodes to identify SAP communities. Spectral methods included the Spectral Mixture Analysis (Ustin et al., 1993; Roberts et al., 1997, 1998; Elmore et al., 2000; Dennison et al., 2004; Lu et al., 2004) and Spectral Angle Mapper (Kruse et al., 1993; Dennison et al., 2004) coupled with reflectance thresholds for specific bands, vegetation red-edge detection – which describes the long-wavelength edge of the Chl absorption, from low reflectance near the maximum absorption to high reflectance at a wavelength where Chl does not absorb energy (Han & Rundquist, 1997; Han, 2002; Bostater et al., 2003) – and the absorption feature of Chla at 680 nm (Han & Rundquist, 1997; Han, 2002). This resulted in a map of SAP throughout the Delta, with a classification accuracy of 80% (for details on the classification method see Hestir et al., 2008).

We then investigated whether there were significant spectral differences between the SAP species in glasshouse spectrometer data and the airborne imaging spectrometer. We used the spectra collected in the glasshouse and, from the imagery, we collected spectra of ‘pure’ SAP end members (> 90% pixel cover of each individual SAP species) based on locations of each species from field measurements. We used ANOVA at each band and principal component analysis (PCA) of all bands to test whether species were separable (Jongman et al., 1995). ANOVA identifies which species are most different and which bands are most important to assess these differences. PCA identifies whether the spectral information locates each SAP species in different regions of the hyperellipsoid created in the data space, and the loadings of the PCA axis indicate the contribution of each of the spectral bands to the final result.

We used discriminant function analysis (DFA) (Im et al., 2008) to predict species identity using both resampled glasshouse spectra and airborne data. Species identity and native status were used as the categories, and reflectance spectra were used as covariates. The class discrimination quality was assessed as correct classification rates for the available field data points: for the native vs nonnative classification, the species conditional correct classification rates, and the overall rate. Where DFA supported high separability of the species, we applied its function to a subset of imaging spectroscopy data to create a map of the distribution of each SAP species where there was high co-occurrence of native and nonnative species after masking the area previously classified as SAP (Hestir et al., 2008).

Pigment concentration

Species differences in Chl concentration often manifest themselves as low reflectance at 440, 480, 650 and 680 nm wavelengths. Unfortunately, utilizing these bands in a radiative transfer model, such as PROSPECT (Jacquemoud & Baret, 1990; Jacquemoud, 1993; Feret et al., 2008), would not result in usable Chl concentration estimates, because of the extremely low reflectance of submerged plants at these wavelengths (J.B. Feret, pers. comm.). Therefore, the interspecies differences in Chl, carotenes and anthocyanin concentrations were tested indirectly using the three band indexes described in Gitelson et al. (2006), using the spectrometer reflectance data (not resampled to HyMap wavelengths and bandwidths). For Chl the best approximation using the green reflectance (550 nm) is estimated as:

image(Eqn 1)

where ρ is the reflectance at the specified wavelengths. We truncated the near infrared (NIR) to be between 760 and 800 nm to match the optimizations described in Gitelson et al. (2006). We also estimated Chl concentration using the red edge, using the following equation:

image(Eqn 2)

For carotenes, we followed the same procedures as for Chl, estimating carotene concentration with both the green- and red-edge reflectance values, using the following equations:

image(Eqn 3)
image(Eqn 4)

Finally, we estimated anthocyanin concentration using the following equation:

image(Eqn 5)

We tested whether Chl, carotene and anthocyanin concentrations were significantly different between species and between native and nonnative species using ANOVA and Tukey’s honestly significant difference tests (Zar, 1999).

Photosynthetic efficiency

Photosynthesis is limited in high-light environments to avoid damage to the photosynthetic reaction center (Grace et al., 2007). During high-light periods, plants are able to activate the xanthophyll cycle to divert excess energy. This cycle consists of two states: at high light de-epoxidation allows conversion of violaxanthin to zeaxanthin via antheraxanthin; when light intensities are reduced, epoxidation reverts zeaxanthin to violaxanthin, via the same intermediary products (Grace et al., 2007). This mechanism is not activated in plants using the C4 pathway, allowing them to maximize photosynthesis in high-light environments (Keeley & Rundel, 2003). In the context of SAPs within a tidal system such as the Delta, having a heterogeneous metabolic functionality for high-light environments (e.g. regular exposure to surface light at low tide) and low-light environments (e.g. regular submersion at high tide in turbid water) would increase fitness. For each species we computed the photochemical reflectance index (PRI; Gamon et al., 1997), which represents a measure of light-use efficiency, and changes in metabolic pathways (Grace et al., 2007). Gamon et al. (1992) showed that PRI values were correlated with the epoxidation state of the xanthophyll cycle in sunflower. PRI is a normalized ratio between reflectance at 531 and 570 nm (Gamon et al., 1997) given by:

image(Eqn 6)

Negative values of larger magnitudes indicate activation of the defense mechanism – photosynthetic inhibition at high-light conditions – while less negative or positive values indicate photosynthesis in high-light conditions (Grace et al., 2007). We used the nonresampled ASD reflectance to test whether PRI values were significantly different between species and between native and nonnative species using ANOVA and Tukey’s honestly significant difference tests (Zar, 1999).

In addition, 10 samples of each of the SAP species to estimate the values of δ13C as a measure of internal carbon concentration were analyzed by the UC Davis stable isotope facility (for details on their standard protocols, see Several confounding factors, however, can affect the measurements of δ13C for aquatic plants, such as the degree to which atmospheric CO2 is in equilibrium with the water mass, input of CO2 from the decomposition of 13C-depleted terrestrial detritus in the water, contribution from the dissolution of 13C-enriched carbonate rock, and seasonal rates of photosynthesis and respiration (Boutton, 1991). The Delta waters are slightly depleted in δ13C (Cloern et al., 2002) compared with terrestrial environments, so the δ13C values are expected to be slightly lower to reflect this depletion. The protocol we used is based on the δ13C of CO2 in the atmosphere. Aquatic plants use two sources of carbon for photosynthesis, uptake of carbonic anhydrase-mediated HCO3 or CO2, but the resulting δ13C signal from either source is likely indistinguishable (Riebesell & Wolf-Gladrow, 1995). All submerged aquatic plants in this study have a CCM, but only a few have a C4-like photosynthetic pathway (Maberly & Madsen, 2002). Thus, we expect that if there are differences between species in the δ13C value, and other factors cannot account for them, these would likely be a result of the photosynthetic pathway (Bowes et al., 2002). Plant species that undergo C4-like photosynthesis are likely to have less negative values of δ13C because they discriminate less against 13C from the photosynthetic substrate as a result of the bicarbonate fixed by phosphoenolpyruvate carboxylation (PEPC) rather than carbon dioxide fixed by Rubisco (Farquhar et al., 1989). For C4 photosynthesis, δ13C values c.−30‰ are expected (Farquhar et al., 1989). However, differences in the range of δ13C values are expected for the aquatic environment, where values from −3 to −50‰ have been observed for aquatic plants (Farquhar et al., 1989; Boutton, 1991; Cloern et al., 2002; Raven et al., 2002).

To assess the consistency between the remote sensing estimates (PRI) and the stable isotopes data, we regressed the estimated PRI values against the δ13C values, using the coefficient of determination (R2) to evaluate how well the two metrics are related and an F-test to assess model fit.


Species differences and separability analysis

Submersed aquatic species had significantly different reflectance in certain regions of the electromagnetic spectrum (< 0.0001; Fig. 2; Table 2) that allowed their identification (PCA and DFA; Fig. 3; Table 3). Both the glasshouse spectrometer data and the airborne imaging spectroscopy data discriminated between natives and nonnatives with 80% certainty, and the individual species’ correct classification rate amounted to 60% (Fig. 4; Supporting Information, Tables S1–S4).

Figure 2.

Average spectral signature for each of the native and nonnative species, and water with the handheld spectrometer (ASD) and the airborne sensor (HyMap). Handheld spectrometer measurements were resampled to the airborne sensor spectral bands. For some species only one measurement was possible. Reflectance is measured in %.

Table 2.   Regions of the electromagnetic spectrum where species were most distinct based on the ANOVA of each of the measured bands by the handheld spectrometer (ASD) and the airborne imaging spectrometer (HyMap).
Scientific namesANOVA
ASD (nm)HyMap (nm)
Egeria densa570–580; 950–1000; 1140–1150550–650
Elodea canadensis700
Ceratophyllum demersum660–670; 900–1050550–650
Stuckenia pectinata600–750; 800–1150Nondifferentiable
Stuckenia filiformisNondifferentiable
Potamogeton nodosus750–1000
Potamogeton crispusNondifferentiableNondifferentiable
Myriophyllum spicatum1030–1050; 1090–1120590–640
Cabomba caroliniana520–560560
Figure 3.

Principal component analysis of handheld spectrometer (ASD) and airborne spectrometer (HyMap) spectra for the submersed aquatic plant species occurring in the Sacramento–San Joaquin River Delta, CA, USA. (a) ASD measurements; (b) HyMap measurements. Native species are represented in green, nonnative species in red, and water in blue. Note that ASD does not include Potamogeton nodosus and turbid water, and that HyMap does not include Stuckenia filiformis, and Elodea canadensis because we did not register pure patches in the field. PCA 1, PCA axis 1; PCA 2, PCA axis 2.

Table 3.   Regions of the spectrum that contributed to the principal component analysis (PCA) and discriminant function analysis (DFA) axis, and the amount of variability explained by each analysis
ASD (nm)HyMap (nm)ASD (nm)HyMap (nm)
Axis 1575–900; 1025–1115710–1200496–511; 619–679; 816505–665
Axis 2425–465452–695557–588; 634; 695–740545–635
Axis 3525–565695–820600–700465; 565
Figure 4.

Area selected to apply the discriminant function scores to the HyMap imagery: (a) hyperspectral imagery overlaid with field data points of patches dominated by natives (green) and nonnatives (red); (b) distribution of Ceratophyllum demersum, Egeria densa, Cabomba carolinensis, Potamogeton crispus, P. nodosus, Myriophyllum spicatum.

The univariate analysis revealed that the species are most separable in the visible (400–700 nm) spectral region (Table 2), as indicated by both PCA axis 2 (Fig. 3) and the discriminant functions (Table 3). Native and nonnative species have different reflectance spectra in the visible region (Fig. 2), with the exception of the invasive M. spicatum. The only SAP species with nonsubmersed leaves, P. nodosus, is spectrally distinct from all other submersed species (Fig. 2).

The HyMap remote sensing data also separate well the natural canopies of native and nonnative species (Fig. 3; Table S3). Many species are rather distinct, including most native species such as C. demersum and P. nodosus; however, nonnative species were the most frequently confounded, especially M. spicatum, Potamogeton crispus and E. densa (Table S2). These three species have distinct spectral signatures (Fig. 2), when acquired with a handheld spectrometer. However, these differences are nearly entirely lost in the airborne spectra (Fig. 2; Tables S1, S2). Fig. 4 represents a section of the Delta where most species co-occur and where we applied the discriminant function to the airborne data. It shows a robust discrimination between native (green) and nonnative (orange-red) species at the pixel level (Fig. 4b; Table S4).

Pigment concentration

Nonnatives had significantly higher concentrations of Chl (Chlgreen: = 24.82, df = 104, < 0.0001; Chlred edge: = 84.24, df = 104, < 0.0001), carotenes (carotenesgreen: = 71.74, df = 104, < 0.0001; carotenesred edge: = 60.61, df = 104, < 0.0001) and anthocyanin (= 6.91, df = 104, = 0.009) when compared with native species (Fig. 5). At the species level there were also significant differences (Chlgreen: = 123.69, df = 104, < 0.0001; Chlred edge: = 188.49, df = 104, < 0.0001; carotenesgreen: = 55.17, df = 104, < 0.0001; carotenesred edge: = 58.01, df = 104, < 0.0001; anthocyanin: = 138.14, df = 104, = 0.009), but similar pigment concentrations grouped native and nonnative species (significantly different groups as letters on top of the box plots in Fig. 5).

Figure 5.

Pigment concentration: chlorophyll (a, b), carotene (c, d), and anthocyanin (e) values for native (CEDE, ELCA, PONO, STFI, and STPE; right panel) and nonnative (CACA, EGDE, MYSP, and POCR; left panel) submerged aquatic plant species in the Sacramento–San Joaquin River delta, CA, USA. Pigment concentrations were significantly different between native and nonnative species, and among species (significantly different species are represented by different letters on top of the box plots, at α = 0.05). The black solid line represents the mean and the box plots are the 95% quantiles. CACA, Cabomba caroliniana; CEDE, Ceratophyllum demersum; EGDE, Egeria densa; ELCA, Elodea canadensis; MYSP, Myriophyllum spicatum; POCR, Potamogeton crispus; PONO, P. nodosus; STFI, Stuckenia filiformis; STPE, S. pectinatus. These indexes are unitless but they were calibrated against pigment concentrations in mg m−2.

Photosynthetic efficiency

We found significant differences in PRI values between species (= 281.64, df = 104, < 0.0001) but not between native and nonnative status (= 0.34, df = 104, = 0.56) with the ASD reflectance data (Fig. 6a). Significantly the highest PRI values (less negative) were observed for E. canadensis and C. caroliniana, followed by P. crispus and E. densa, then by S. filliformis, C. demersum and M. spicatum, and the lowest PRI values (more negative) were for S. pectinata.

Figure 6.

Photochemical reflectance index (PRI; unitless) (a) and carbon isotope ratio (δ13C; in & fractions) (b) values for native (CEDE, ELCA, PONO, STFI and STPE; left panel) and nonnative species (CACA, EGDE, MYSP, and POCR; right panel). (c) PRI and δ13C regression analysis for native (closed circles) and nonnative (open circles) submerged aquatic plant species in the Sacramento–San Joaquin River delta. Values of PRI (unitless) and δ13C (in ‰ fractions) were significantly different among species (significantly different species are represented by different letters on top of the box plots, at α = 0.05). The black solid line represents the mean and the box plots are the 95% quantiles. CACA, Cabomba caroliniana; CEDE, Ceratophyllum demersum; EGDE, Egeria densa; ELCA, Elodea canadensis; MYSP, Myriophyllum spicatum; POCR, Potamogeton crispus; PONO, P. nodosus; STFI, Stuckenia filiformis; STPE, S. pectinatus.

Stable isotope results were consistent with those of PRI, especially for the nonnatives E. densa, M. spicatum, P. crispus, and the native C. demersum. Nonnative species showed significantly less negative δ13C values than the natives (= 27.52, df = 89, < 0.0001), with the exception of the nonnative C. caroliniana (the most negative) and the native Stuckenia spp. (the least negative; Fig. 6b). C. demersum and E. canadensis showed the greatest range of variation in isotope ratio. We were unable to compare isotope ratios and PRI for P. nodosus because samples for this species were unavailable at the time of measurement. PRI and δ13C values were strongly related for both native and nonnative species (Fig. 6c). Regression analysis coefficient of determination (R2) was 0.41 for both native and nonnatives, and the model was significantly fitted to the data (natives: = 25.9, = 0.0001; nonnatives: = 26.2, = 0.0001; Fig. 6c).


Our results show that morphological differences in plant structure and biochemistry allow spectral differentiation between natives and nonnatives. While many previous studies succeeded in discriminating terrestrial species (Cochrane, 2000; Lewis, 2000; Fyfe, 2003; Andrew & Ustin, 2006; Hutto et al., 2006; Atkinson et al., 2007), the aquatic environment presents a much greater challenge to spectral differentiation (Marshall & Lee, 1994; Hestir et al., 2008). So far, only a few submerged aquatic species have been successfully differentiated (Williams et al., 2003; Dogan et al., 2009). This differentiation challenge results from linear and nonlinear mixing with water, as well as the properties of the water column itself (Williams et al., 2003; Hestir et al., 2008; Dogan et al., 2009). Our data collection was designed to reduce the known challenges of optical remote sensing over aquatic systems, which include meteorological and illumination variability, in-water radiance, and water-leaving radiance (Giardino & Zilioli, 2001; Holden & LeDrew, 2001; Bostater et al., 2003; Vis et al., 2003; Williams et al., 2003; Dogan et al., 2009). To account for meteorological and illumination variability, we specified flight times that minimized clouds and specular reflectance and we controlled for wind velocities and time of the day during data acquisition (Hestir et al., 2008). In-water radiance varies with the ratio of SAP to water in the water column, as different mixtures have different results in the amount of radiance. In the Delta there are freshwater outflows with high turbidity gradients, which mix with tidal water inflows. These dynamics make the in-water radiance highly variable during the daily tidal cycles. To reduce the effects of in-water radiance variability we restricted our analysis to pure pixels, where individual SAP species were the dominant cover in the pixel rather than water (Hestir et al., 2008). Finally, SAP contribution to water-leaving radiance is likely affected by the depth of the water column above the SAP cover (Han & Rundquist, 1997; Han, 2002). To reduce this effect we restricted our imagery collection to low-tide conditions (Hestir et al., 2008). In a parallel study testing the impact of the water column, we observed no significant deterioration of the submerged plant spectra with depth, and the water column overlying the canopy did not limit the plant detectability (Hestir, 2010).

We have shown that native and nonnative species have systematic differences in their spectral properties related to biochemistry, light use, and morphological, and structural traits, in both ‘controlled’ and natural canopies. At the leaf level, reflectance is affected by the structure of the leaf tissue and biochemistry, and the size, shape, and orientation of the leaf. Submerged leaves have a poorly differentiated mesophyll and a high frequency of epidermal chloroplasts that are likely to be under selective pressure by the reduced diffusion coefficient of carbon dioxide in water (MacFarlane & Raven, 1990), which is not offset by the use of bicarbonate in photosynthesis (Raven et al., 2005). The presence of epidermal chloroplasts that maximize light absorption by the submersed leaf (Sculthorpe, 1965) results in minimal spectral differences between species. Hence, we believe that the structure of leaf tissue may be more important to differentiate between nonsubmersed and submersed plants rather than between co-occurring submersed plants. Nonsubmersed aquatic plants have a greatly differentiated mesophyll with palisade and spongy layers, and internal anatomy similar to land plants. In fact, our results show that the only SAP species with nonsubmersed leaves, P. nodosus, is spectrally distinct from all other submersed species (Fig. 2). The leaf structure of S. pectinata is also substantially different from the other species – stems without leaf blades, allowing S. pectinata to be spectrally separated from other SAPs in the glasshouse spectrometer data (Table 2). In the landscape, however, S. pectinata occurs in very sparse canopies, which often form patches smaller than the ground pixel size of 9 m2, producing low biomass per pixel area. As a result, S. pectinata is indistinguishable from other species in the airborne data (Table 2), despite its very characteristic laboratory spectrum (Fig. 2).

Plant leaf biochemistry greatly influences reflectance (Ustin et al., 1991, 2009; Ustin et al., 2004). This factor is not totally independent of the internal leaf arrangement, as leaf optical properties often correspond to pigment concentrations (Blackburn, 2007; Ustin et al., 2009). Chlorophylls, carotenoids, and other pigments have absorption peaks at overlapping but different wavelengths (Ustin et al., 2009), between 400 and 700 nm, which were used to separate species (Table 3). Several plant pigments are instrumental in plant photosynthetic activity (Ustin et al., 2009), and differences in leaf biochemistry affect photosynthetic efficiency in moles of carbon assimilated per mole of photons absorbed.

Different leaf widths, shapes, and colors of these species may also contribute to the measured reflectance. With the exception of P. nodosus, most native species have no leaf blades, dissected leaf blades (e.g. C. demersum) or narrow blades (e.g. E. canadensis), which are distinctly different leaf morphologies from the wide blades and large dissected leaf whorls of nonnative species. These growth forms may influence the amount of light intercepted, and thus the light reflected. Our analysis showed that native species are spectrally distinct from nonnatives, which is likely a result of three convergent characteristics of nonnative species: wider ribbon-like leaves, greater leaf area per plant with higher Chl concentration, resulting in lower visible reflectance. Native and nonnative species have different reflectance in the visible region and most of the nonnative species have darker green leaves while native species are a brownish color, indicating different pigment compositions (Ustin et al., 2009). Our results corroborate this prediction as we found significantly higher pigment concentrations (Chl, carotenes and anthocyanins) in nonnative than in native species. Larger differences were found for pigments that harvest photons for photosynthesis (Chl and carotenes) than for anthocyanins. Our spectral profiles show significant differences in reflectance that match these differences in pigment concentrations. The PRI results show that potentially different pigment concentrations are present in native and nonnative species (see paragraph on 13C and PRI results). We conclude that the observed reflectance patterns are related to the interaction of shape, width and color of the leaves, which explains the separability of these species.

When scaling up from the leaf to the canopy, factors such as leaf density and canopy closure come into play. Many species are quite distinct, including most native species, such as C. demersum and P. nodosus; however, the ones most frequently confounded were the nonnative species, especially M. spicatum, P. crispus and E. densa. The canopy of the nonnative species found in the Delta tend to have higher leaf density (Fig. 1), resulting in high reflectance in the NIR (Fig. 2), and the spectral signatures are less impacted by the surrounding water. Thus, the effect of water absorption in this part of the spectrum should be less evident than measurements of canopies of native species. Furthermore, the effect of the water column as a potential confounding factor to the discriminant analysis was minimal because the imagery was acquired to avoid specular reflectance and at low tide, when most of the submersed plant canopies are at or near the surface.

Myriophyllum spicatum is overclassified (Fig. 4) and is often confused with E. densa and P. crispus, contrasting with their current Delta-wide distribution. E. densa is ubiquitous (Hestir et al., 2008), and tends to exist in most channels, in areas of moderate- and low-velocity water, shallow and deep waters, turbid and clear waters, and water with variable salinity (Santos et al., 2011). M. spicatum tends to be more restricted to somewhat deeper and more turbid waters with higher salinity (Grace et al., 2007), in the western part of the Delta. The great range of environmental conditions and species assemblages required to develop a classified map for the Delta may have washed out site-specific spectral differences between the species in the area represented in Fig. 4. While training the classifier was done to encompass the variability at the larger scale, it may have led to misclassification of the species at this finer scale. Additionally, in this region E. densa was frequently associated with epiphytic algae that could have contributed to its misclassification. In fact, the patches classified as pure E. densa did not have algal growth, while those with algal growth were misclassified.

Our results show distinct δ13C and PRI values between native and nonnative species which were less negative for nonnative than for native species, and the two metrics were strongly correlated. Our results are within the ranges of other published δ13C (Cloern et al., 2002) and PRI data (Peñuelas et al., 1993). Several confounding factors could affect the interpretation of δ13C values, especially if native and nonnative species experienced different aquatic environments. In the Delta, natives and nonnatives co-occur throughout their distribution range (Cohen & Carlton, 1995; Jassby & Cloern, 2000; Lucas et al., 2002; Santos et al., 2011), all of which occupy slower water channels, where submerged plants are associated with a Delta-wide decrease in turbidity (Hestir, 2010), through their effects on sedimentation processes. All but one native species have roots, as do all the nonnatives (Table 1), and all experience δ13C from the same water sources. Thus we believe that there is a low probability that natives and nonnatives are experiencing the δ13C environment differentially, suggesting that the observed differences are the result of physiological differences. One possible explanation for these patterns is that different CCMs result in different δ13C signals. Since most native and nonnative SAPs have CCMs (Maberly & Madsen, 2002) and the resulting δ13C signal from either HCO3 or CO2 uptake is likely indistinguishable (Riebesell & Wolf-Gladrow, 1995), we can discard this possibility as explaining the observed differences. Alternatively, nonnative species have both C3 and C4-like photosynthetic pathways (Van et al., 1976; Salvucci & Bowes, 1981; Casati et al., 2000; Maberly & Madsen, 2002), which can overcome the high-light and -temperature limitations of C3-only plants. This may give nonnatives (mostly E. densa, and M. spicatum) the ability to maintain photosynthesis under high light and high temperature, which is supported by previous studies that have demonstrated that E. densa maintains continuous growth throughout the year (Pennington & Sytsma, 2005; Pennington, 2007; Santos et al., 2011). Our results also showed less negative PRI and δ13C values for P. crispus, which may indicate the presence of a C4-like mechanism, as suggested in previous research (Sand-Jensen, 1983; Nichols & Shaw, 1986). Exceptions, however, occur for E. canadensis and C. caroliniana, which have the highest PRI values and among the lowest δ13C values, suggesting both C3 and C4-like mechanisms as shown in previous research (Salvucci & Bowes, 1981; Sand-Jensen, 1983), or in the case of C. caroliniana that high-light photosynthesis can be activated with CO2 as a substrate (Smith, 1937). Finally, Stuckenia spp. showed the lowest PRI and the highest δ13C values, potentially because these species are heterophyllous, and their canopies are formed by submerged, emergent and terrestrial leaves that have high heterogeneity in photosynthetic traits (Iida et al., 2009).

The adaptive value of having a facultative C4-like photosynthesis allows plants to colonize environments that C3-only plants cannot utilize or utilize less efficiently, such as the high-light and -temperature conditions of shallower waters, and tidal sites of increased salinity (Nichols & Shaw, 1986). These conditions are common throughout the Delta, with the exception of water with carbon (and nutrient) limitations (Jassby & Cloern, 2000; Lucas et al., 2002). This plasticity in traits may allow the nonnatives to persist and succeed in the new environment (Simberloff & Holle, 1999; Simberloff, 2001). The nonnative species can occupy the same environments in which the native species occur, but also environments where natives are not competitive, giving nonnatives an advantage by occupying a wider range of environments of the Delta. Even if remote sensing identification is limited to identifying native and nonnative submersed species, this work advances systematic measurements of specific traits that may contribute to understanding invasibility and invasion success. This is a new approach to study invasiveness and elucidate why some species are more competitive, and we hope this information will be incorporated into systems for early detection and monitoring of new and recently introduced species.


Funding for this research was provided by the California Department of Boating and Waterways Agreement 03-105-114, and the California Department of Water Resources Contract #4600008137 T4. We would like to acknowledge D. Kratville, J.R.C. Leavitt, P. Akers, and the field crews of the California Department of Boating & Waterways; and J. Greenberg, M. Andrew, P. Haverkamp, M. Whiting, and A. Kultonov for useful discussion and sharing of ideas on environmental conditions in the Delta. We also thank R. McIlvaine and G. Scheer for administrative support.