Plant spectral diversity — how plants differentially interact with solar radiation — is an integrator of plant chemical, structural, and taxonomic diversity that can be remotely sensed. We propose to measure spectral diversity as spectral variance, which allows the partitioning of the spectral diversity of a region, called spectral gamma (γ) diversity, into additive alpha (α; within communities) and beta (β; among communities) components. Our method calculates the contributions of individual bands or spectral features to spectral γ-, β-, and α-diversity, as well as the contributions of individual plant communities to spectral diversity. We present two case studies illustrating how our approach can identify “hotspots” of spectral α-diversity within a region, and discover spectrally unique areas that contribute strongly to β-diversity. Partitioning spectral diversity and mapping its spatial components has many applications for conservation since high local diversity and distinctiveness in composition are two key criteria used to determine the ecological value of ecosystems.

Leaf spectra

Leaf spectra were measured with a portable field spectrometer (ASD FieldSpec 4, Malvern Panalytical, Cambridge, UK), covering the wavelength range from 350 nm to 2500 nm and an integrating sphere with internal light source (ASD RTS-3ZC, Malvern Panalytical, Cambridge, UK) in the summer of 2017, following the leaf spectroscopy protocol from the Carnegie Spectranomics project (https://cao.carnegiescience.edu/spectranomics-protocols), with some modifications (Laliberté & Soffer 2018). The tree species sampled (with five individual plants per species, each representing the average spectrum from six mature leaves) were Betula alleghaniensis Britton (yellow birch), Populus deltoides W. Bartram ex Marshall subsp. deltoides Marsh (eastern cottonwood), and Populus tremuloides Michaux (trembling aspen). Processing of spectra consisted of applying a third-order Savitzky-Golay filter (length = 55) to reduce noise, reducing spectral resolution from 1 nm to 10 nm wide to reduce the number of bands, trimming the spectra between 410 and 2400 nm to remove regions with low signal-to-noise, and brightness-normalizing spectra. This vector normalization emphasizes differences in the shape of spectra as opposed to differences in amplitude (i.e. albedo or brightness). The R code to perform the analyses is available online (https://github.com/elaliberte/specdiv).

NEON imaging spectroscopy data

Second, we applied our method for partitioning spectral diversity to imaging spectroscopy data collected by NEON’s Airborne Observation Platform (AOP; Kampe et al. 2010) over the Bartlett Experimental Forest (https://www.neonscience.org/field-sites/field-sites-map/BART). The AOP Data Product used was NEON_D01_BART_DP3_314000_4880000_reflectance.h5, and was downloaded from http://data.neonscience.org on 25 February 2018. In this case study, we used a scene measuring 280 m (east-west) x 1000 m (north-south), acquired in August 2017. Spectral data were processed to surface reflectance and subsampled to 1-m pixel size by NEON. The R code to perform the analyses is available online (https://github.com/elaliberte/specdiv).

References

Kampe, T.U., Johnson, B.R., Kuester, M. & Keller, M. (2010). NEON: the first continental-scale ecological observatory with airborne remote sensing of vegetation canopy biochemistry and structure. J. Appl. Remote Sens., 4, 043510-043510–24.

Laliberté, E. & Soffer, R. (2018). Measuring spectral reflectance and transmittance (350-2500 nm) of large leaves using the Spectra Vista Corporation (SVC) DC-R/T Integrating Sphere. protocols.io. https://dx.doi.org/10.17504/protocols.io.p8pdrvn