We report a recently developed application of airborne imaging spectroscopy analyzing ecosystem changes caused by invasion. Remote sensing has been used to map the distribution of some biological invaders (4, 5), but invasions often must be far advanced before they can be detected by conventional remote sensing that relies on canopy emergence and/or dominance by the invader (6). In any case, conventional approaches yield the distribution of the invader, not the ecosystem-level effects of invasion. Our approach uses imaging spectroscopy to evaluate changes in canopy chemistry and other canopy characteristics caused by invasion. Canopy chemistry (particularly nitrogen concentration) is causally connected to and highly correlated with rates of plant productivity and nutrient use efficiency, decomposition, and nutrient availability in soil (7–9). Consequently, changes in canopy chemistry can be used to show not just where invasion occurs, but also whether it has a substantial effect on ecosystem-level biogeochemistry.

Efforts to measure canopy chemistry by remote sensing have been underway for some time (10–15), and have yielded useful empirically based estimates. Recent advances in theory, notably spectroscopic transport models that trace photons through the canopy (16–19), and in the development of airborne sensors with signal-to-noise, stability, and dynamic range performances matching laboratory spectrometers (20, 21) allow for assessment of multiple biogeochemical properties simultaneously across a landscape.

We applied this approach to the invasion by the Canary Islands tree Myrica faya into tropical montane forests of Hawaii Volcanoes National Park (HAVO). HAVO is useful for this research because most of the montane forest area of the park is dominated by a single overstory tree species, Metrosideros polymorpha, providing a consistent background against which we can evaluate change; and because nitrogen (N) limits forest productivity and causes low foliar N concentrations in this young volcanic landscape (9, 22). Myrica is a symbiotic N₂ fixer, a capacity that is lacking in native trees of these young volcanic sites. Its leaves are much richer in N than Metrosideros, and invasion by Myrica increases N input and availability severalfold, profoundly altering the pathway of ecosystem development in the sites it dominates (1, 23, 24).

We used the recently upgraded National Aeronautics and Space Administration Airborne Visible and Infrared Imaging Spectrometer from an ER-2 high-altitude aircraft to measure canopy fractional cover, canopy water content, and leaf N concentrations in a 1,360-hectare area near the summit of Kilauea Volcano in HAVO (Fig. 1) (see Supporting Text and Figs. 5–7, which are published as supporting information on the PNAS web site). These remote measurements were complemented by extensive ground-based structural, spectroscopic, and chemical analyses of forest canopies. Native forests in the region are dominated by Metrosideros, with the tree fern Cibotium glaucum common in the forest understory. Ground-based measurements show that Metrosideros forests in this area are 10–20 m tall, with relatively high leaf area index (LAI = 4–6), but low leaf N (0.6–0.8%) and water (H₂O; 45–55%) concentrations. Where it occurs, the Cibotium understory is 3–10 m tall; it has low LAI (1–2) but relatively high leaf N (1.2–1.6%) and moderate H₂O (58–67%) concentrations. In contrast, Myrica supports very high LAI (6–11) and high leaf N (1.5–1.8%), but moderate leaf H₂O (50–65%) concentrations in the stands it dominates.

Fig. 1. Southeast flank of Kilauea Volcano in Hawaii Volcanoes National Park, showing recent volcanic activity together with dense tropical forest and its transition to more open woodlands, pastures, and other forms of land use. Although most of this area is (more ...)

Remote analyses of upper-canopy foliar N concentrations correlated strongly with ground-based measurements (Fig. 2A, slope = 0.93, r² = 0.91), in the 47 image pixels for which we had information. This correlation held up within Metrosideros-dominated and Myrica-dominated canopies, as well as between them. Similarly, remotely measured canopy water contents correlated highly with ground-based measurements (Fig. 2B, slope = 0.90, r² = 0.92).

Fig. 2. Comparison of remotely sensed estimates of leaf N concentration and canopy H2O content with ground-based sampling and laboratory analysis. Areas dominated by Metrosideros, Myrica, mixed Metrosideros–Hedychium, and mixed Metrosideros–Cibotium (more ...)

Aircraft-based analyses of the HAVO landscape yielded strong spatial gradients in both leaf N concentration and canopy water content (Fig. 3). Some areas contained both high foliar N and canopy H₂O, whereas others were a mix of high or low values, and still other areas were low in both H₂O and N (Fig. 6). Extensive field observations demonstrated that low canopy H₂O and leaf N concentrations indicated dominance of native Metrosideros trees, whereas high canopy H₂O and leaf N concentrations were associated with well established stands of invasive Myrica. Most Myrica-dominated stands are surrounded by areas with intermediate foliar N and slightly elevated canopy water content (Fig. 3); in these areas, individual Myrica trees are reaching into the canopy but do not yet dominate it, meaning that we are detecting the spread of the invader by its effect on canopy chemistry before it dominates the forest.

Fig. 3. Leaf N concentrations and canopy H2O content estimated at 9 × 9m spatial resolution in a 1,360-hectare area of Hawaii Volcanoes National Park, using airborne high-fidelity imaging spectroscopy and photon transport modeling (also see supporting (more ...)

We were surprised that the airborne measurements also identified substantial areas with low foliar N but relatively high canopy water content (Fig. 3), and still more surprised to discover that, when we visited those areas, their understories were often dominated by the invasive tall herb Hedychium gardnerianum (Kahili ginger). Hedychium is known to be a widespread invader of Hawaiian forests, in HAVO and elsewhere (25), but because it colonizes the forest understory, it is invisible to conventional remote sensing techniques. Its canopy differs substantially from those of Metrosideros and Myrica; although it reaches only 1–2 m in height in the forest understory, it supports relatively high LAI (3–5), and high foliar N (1.8–2.1%) and water (80–93%) concentrations. Because analysis of canopy water content by airborne imaging spectroscopy measures the returning photons that interact with the entire volume of plant tissues in the pixel (see supporting information), it allows the detection of Hedychium under the forest canopy. In contrast, airborne imaging spectroscopy is sensitive to foliar N only in the upper canopy, and so the high N content of understory Hedychium is undetectable.

In fact, Metrosideros stands that have been invaded by Hedychium appear to have lower foliar N than uninvaded stands, a pattern that we first detected remotely, and later confirmed with ground-based sampling (Fig. 2A). We believe that this decrease in foliar N in Metrosideros is due to N uptake by nutrient-demanding Hedychium; the alternative, that Hedychium preferentially invades nutrient-poor sites, is contradicted by observations that its invasion responds strongly and positively to increased nutrient availability (26). In addition to its biogeochemical effects, the dense shade and network of tubers and roots that Hedychium forms in invaded sites serves as an effective barrier to the establishment of native plant species (25).

Regional statistics for the Metrosideros, Myrica, and Metrosideros–Hedychium classes demonstrate that invasion significantly alters canopy chemistry at the landscape scale (Fig. 4). Kolmogorov–Smirnov tests showed that all three forest types differed in foliar N distributions (P < 0.05). The Metrosideros–Cibotium canopy distribution is negatively skewed toward relatively low values (0.68 ± 0.11%), whereas Myrica stands have a positively skewed distribution with much higher N concentration values (1.56 ± 0.32%). Hedychium invasion is spatially linked to decreased foliar N concentrations of overstory Metrosideros (0.51 ± 0.14%). Moreover, Hedychium increases aboveground water content by nearly 50% over native Metrosideros stands, whereas Myrica more than doubles canopy water (Fig. 4).

Fig. 4. Cumulative histograms of leaf N concentration and canopy H2O content values from imaging spectrometer measurements of areas dominated by: native Metrosideros forest, invasive N-fixing Myrica, and invasive Hedychium in Metrosideros understory (also see (more ...)

Classification of the area based on canopy chemistry (Fig. 6) suggests that ≈28% of the landscape is dominated by Myrica, and that an additional 23% is undergoing transformation as Myrica grows into the canopy. Up to 13% of the remaining Metrosideros forest is infested with Hedychium, directly altering biogeochemical conditions at the understory level and indirectly at the overstory forest via nutrient competition. Assuming that all current forested areas were once dominated by Metrosideros, invasion of Myrica has increased canopy N >4-fold locally, and approximately doubled canopy N content over the 1,360-hectare area as a whole.§

These observed changes in canopy chemistry are linked to several fundamental changes in ecosystem properties. Higher foliar N in Myrica canopies is associated with faster leaf turnover, higher rates of nutrient cycling, faster decomposition, greater N availability, greater fluxes of N-containing trace gases, and ultimately invasion by other nutrient-demanding species (23, 27, 28). For example, nitrification rates in soil increase from near zero in Metrosideros stands (–0.04 ± 0.01 mg of N per g of soil per day) to 1.3 ± 2.3 in areas containing Myrica saplings, and to 5.2 ± 3.7 in areas dominated by Myrica (see supporting information). The biogeochemical consequences of invasion by Hedychium have not been evaluated as thoroughly because this invasion's effect on overstory trees had not been identified previously. High-N Hedychium foliage decomposes much more rapidly than does Metrosideros (29), which could enhance overall rates of nutrient cycling (30), but the decrease in Metrosideros foliar N could offset some or all of this effect.

Airborne imaging spectroscopy can provide a unique, spatially detailed understanding of the biogeochemical impacts of different species, including invaders, throughout a forest landscape. Indeed, the chemical, structural, and biological diversity of rain forests in particular requires wholly new remote sensing approaches to understand spatial and temporal variations in nutrient, water, and carbon cycle processes. These airborne observations and techniques cannot substitute for ground-based biogeochemical studies, but they can direct field measurements based on a regional understanding of canopy chemistry.

We thank H. Farrington, R. Martin, D. Turner, A. Warner, K. Carlson, K. Lo, A. Chin, M. Meyer, K. Amatangelo, and S. Robinson for assistance with field and laboratory analyses, and C. Field, R. Martin, S. Ollinger, S. Porder, A. Townsend, and S. Ustin for helpful comments on the manuscript. This work was supported by the National Science Foundation Grants DEB-0136957 and DEB-0108492, the National Aeronautics and Space Administration New Investigator Program (NAG5-8709), The Andrew Mellon Foundation, and The Carnegie Institution.