Methods

2.1. Study Area

The investigation was carried out in a wetland complex covering approximately 10 km2 in the lower Muskegon River Watershed (MRW) located in west-central Lower Michigan, USA (W86° 09' 45", N43° 16' 10"). Fed from the Muskegon River and an extensive tributary network, this wetland complex serves as the last land-water interface before draining into Lake Muskegon, which then flows directly into Lake Michigan. Lake Muskegon (1,679 ha) was recognized as an Area Of Concern by the 1987 Great Lakes Water Quality Agreement due to declining aquatic ecosystem conditions. The National Wetland Inventory classifies the majority of the wetland complex as palustrine containing seasonally and semi-permanently flooded regions of scrub-shrub, forest, and emergent covers. Land uses adjacent to the wetland complex include residential and commercial neighborhoods, urban parks, industrial zones including a pulp and paper mill, chemical and petrochemical facilities, and agricultural and forest patches.

2.2. Data Collection and Preprocessing

A field campaign was conducted during mid-August (2006), which generally represents the peak of the growing season for wetland vegetation within the study area. Capturing data during peak phenological cycles has been shown to increase the separability of invasive wetland plants [25]. Due to the expansive nature of the wetland complex and the challenge of moving through a wetland, a compromise between operational feasibility and statistical sampling rigor was required. It was not the objective of this research to collect an extensive spectral library capturing the phenological and/or regional spectral differences displayed by common wetland plants. Rather, it was our intent to test the applicability of three processing techniques on a small but well controlled number of plant spectra. Both logistical constraints (equipment setup and takedown) and traveling throughout the wetland complex required a substantial amount of time. Reconnaissance field work identified emergent pools where high biodiversity and ecologically noteworthy species of interest (i.e., invasives) were present. Focusing our efforts around these regions of the wetland complex allowed a representative set of species spectra to be collected. An airboat provided the most efficient access for traveling around the wetland complex. Data acquisition focused on the dominant terrestrial-, emergent-, and submergent- species. Dominance was qualitatively identified during reconnaissance field work by evaluating percent cover and the approximate size of a patch for a species. A total of twenty-two wetland plant species were recorded (Table 1). Eight species are identified as being invasive [26]. Note that not all the species identified as invasive are classified as exotic and the degree of 'invasiveness' can vary by region and conditions.

Table 1. Plant species sampled in the study area

Scientific name

Common Name

Invasive

Sagittaria latifolia

Arrowhead, broadleaf

Invasive

Scirpus validus

Bulrush, softstem

Typha latifolia

Cattail, broad-leaved

Invasive

Leersia oryzoides

Cutgrass

Vallisneria americana

Eelgrass

Eleocharis rostellata

Spikerush, beaked

Elodea canadensis

Canadian waterweed

Invasive

Sparganium androcladum

Bur-reed, branched

Filamentous green algae

Heteranthera dubia

Water star grass

Iris versicolor

Iris, harlequin blue flag

Lemna minor

Common duckweed

Myriophyllum verticillatum

Watermilfoil, whorl-leaf

Mowed field grass

Polygonum pensylvanicum

Pennsylvania smartweed

Invasive

Phragmites australis

Common reed

Invasive

Pontederia cordata

Pickerelweed

Invasive

Potamogeton spirillus

Spiral pondweed

Lythrum salicaria

Purple loosestrife

Invasive

Nymphaea odorata

White water lily

Invasive

Salix eriocephala

Willow

Nuphar lutea

Yellow pond lily

We used a portable spectroradiometer (FieldSpec Pro FR®, Analytical Spectral Devices, Inc., Boulder, Colorado) to collect in situ radiance between 350-2500 nm (visible to shortwave infrared). Spectral resolution (full width half maximum) was approximately 3 nm in the visible wavelengths and 10 nm in the infrared region. The sensor was equipped with a 24 degree field-of-view (FOV) optic and held approximately 1-meter above the target at nadir for measurements representing field-canopy conditions. Sun-target-sensor geometry was repeated as best as possible under these difficult field conditions between 11:00-14:30 local time. This viewing geometry configuration approximately represents the spatial resolution that current airborne hyperspectral sensors can achieve (-approximately 1m). A Spectralon® panel (Labsphere, Inc., North Sutton, New Hampshire) was used for calibration during processing and atmospheric adjustments. Eight rapidly sequenced measurements were averaged for one spectrum and eight or nine spectra (depending on the amount of non-overlapping FOVs within a patch) were collected in a homogeneous (one species) plot. The instrument was shifted within the patch during collection to capture inherent within patch variability and ensure non-overlapping FOVs. This was repeated for four patches for each plant. This process provided 32 unique measurements for each species. During data acquisition, the sensor was first placed over the reference panel to record the panel-reflected radiance. Then the sensor was placed over the target to record the target-reflected radiance. Then, by ratioing the radiance measurements, the surface reflectance factor was calculated. By definition, the term reflectance factor is the ratio of radiant emittance of a target (i.e., wetland plant) to that reflected into the same reflected-beam geometry and wavelength range by an ideal and diffuse standard surface (i.e., the Spectralon panel) irradiated under the same conditions [27]. The reflectance factor was calculated based on the following equation:

where p is in situ reflectance factor for target of interest (wetland plant species), tE T is target (wetland plant species) in situ radiance, and cpE T is the calibration panel in situ radiance.

Subsequent data processing in this study also removed wavelength regions severely affected by atmospheric absorption in the spectral ranges of 1350-1480, 1775-2000, and >2400 nm [22]. Figure 1 displays invasive spectra.

2.3. Analytical Techniques

A variety of measures of separation exist that quantify the degree of dissimilarity between any two probability distributions. One commonly applied separation measure is the Jeffries-Matusita (JM) distance measure. A few versions of the JM distance equation exist throughout the remote sensing literature [10, 28]. For the purposes of this analysis, the JM distance equation presented by Niel et al. [28] was utilized. In this version, a JM distance value can range from 0 (i.e., identical distributions) to 1.414 (i.e., complete dissimilarity). We used the JM distance measure (Eq. 2 & 3) to evaluate spectral separability and assess the process of continuum removal. :

where JMij is the Jeffries-Matusita distance between signatures i and j and a is the Bhattacharyya distance.

Ci+Ci 2

VCWiI

where i and j represent the two classes of interest, T is transpose, C. is the variance-covariance matrix of signature i, U is the mean vector of signature i, | C. | is the determinant of C

Phragmites australis

x*x*" \

*

^ X

XX** ***

0 T3^-1-1-1-1-1-1-1-1-1-1-1-1-1-1-1-1-1-

Pontederia cordata

!

^rJ

Figure 1. Reflectance factor at 50nm intervals for selected invasive plant species. Pre-processing removed wavelength regions severely affected by atmospheric absorption in the spectral ranges of 1350-1480, 1775-2000, and >2400 nm. Averaged spectra (n=32).

A modified continuum removal technique was developed for the wetland vegetation spectra following methods outlined in Schmidt and Skidmore [10]. The modification was necessary because some reflectance peaks can disregard relatively smaller, yet potentially important reflectance peaks. For example, the reflectance peak associated with biomass for green vegetation can have reflectance in the near-infrared an order of magnitude higher than that found in the visible portion of the spectrum. Therefore, a piecewise or 'modified continuum removal' was applied to the plant spectra. Specifically, a modified convex hull was forced, or fit, to include seven primary spectral reflectance maxima distributed among separated spectral regions. Once the modified continuum removal was fit, the continuum was removed by dividing the reflectance by the convex hull [16]. The seven primary spectral regions isolate the major reflectance features and were:

• Visible domain and chlorophyll absorption region (350-675 nm)

• Near-infrared plateau (781-975 nm)

• Near-infrared down slope (976-1190 nm)

• Upper near-infrared shoulder (1191-1450 nm)

• First shortwave infrared plateau (1451-2000 nm)

• Second shortwave infrared plateau (2001-2400 nm)

The second derivative analysis applied in this research was conducted to characterize potentially unique wavelength locations of absorption and reflection features within the collected spectra. In this research, a MATLAB script was created that fit a piecewise cubic spline to smooth a non-continuous/unsmoothed spectra in order to create a polynomial from which true second derivative values could be calculated at each band location. The five highest magnitude positive and negative values were selected to identify wavelengths possessing distinct diagnostic spectral change. This percentage was chosen because inspection of the data shows that derivatives and their paired wavelengths resulting from inflection points caused by system noise and not botanical sources were more frequent in the "middle" 80% of the data. The high magnitude values represent points of inflection that are located at the center of a reflectance (negative values capture convex features) and/or absorption feature (positive values capture concave features).

The third identification approach utilizing vegetation reflectance variation was developed by Cochrane [23] for distinguishing tropical tree species. This approach uses a shape-filter representing the range of reflectance variation present in a species over the spectrum. The maximum and minimum spectral reflectance for a given species for each wavelength creates its shape-space. If any of the comparison spectra (e.g., invasive species of interest vs. all others) fall outside the shape-space, then separation is possible. If spectra overlap for a given wavelength, then separation at that wavelength is not possible. This shape filter [23] was applied to evaluate separability of wetland invasive species in the study area. The maximum (Max) and minimum (Min) spectral reflectance at each band center/wavelength creates the shape-space for the species (Eq. 4).

0 0

Post a comment