Linking imaging spectroscopy and LiDAR with floristic composition and forest structure in Panama
Introduction
Imaging spectroscopy and LiDAR (Light Detection and Ranging) have become powerful tools in the study of ecosystems. Unlike multispectral sensors such as Landsat or the Moderate Resolution Imaging Spectroradiometer (MODIS), which integrate reflectance in visible to shortwave wavelengths over a relatively small number of broad-bandwidth spectral channels, imaging spectroscopy measures reflected electromagnetic radiation on a continuous narrow-band (e.g. 10 nm) basis. This provides spectra for detailed analysis of remotely-sensed features and hence diagnosis of their chemical properties (Kokaly et al., 2009, Ustin et al., 2009). From airborne data, these features may resolve forest plots or individual tree crowns. LiDAR data are equally useful for studying ecosystems. LiDAR is an active sensing system in which return times for laser emissions are used to calculate the elevation of both terrain and aboveground features. In forested ecosystems, LiDAR returns can penetrate the canopy and provide information from both the ground and multiple points within the forest canopy (Dubayah et al., 2010). These data can then be used to generate high-resolution images of forest structure, allowing estimates of forest properties such as canopy height and biomass (Drake et al., 2002, Lefsky et al., 2002).
Imaging spectroscopy and LiDAR data have been used in a variety of ecosystems to study vegetation composition, structure, and dynamics. Both spectroscopy and LiDAR data have individually been used to predict species richness in temperate forests (Leutner et al., 2012), tropical forests (Carlson et al., 2007, Kalacska et al., 2007, Wolf et al., 2012), savannas (Cho et al., 2012), and tropical mangrove ecosystems (Held, Ticehurst, Lymburner, & Williams, 2003). In addition, imaging spectroscopy has been used to identify changes in species composition in savanna (Baldeck & Asner, 2013), temperate grassland (Schmidtlein and Sassin, 2004, Schmidtlein et al., 2007), and mixed forest/non-forest ecosystems (Leutner et al., 2012); and imaging spectroscopy has been used to map the distributions of individual species in tropical forests and savannas (Asner and Vitousek, 2005, Clark et al., 2005, Feret and Asner, 2012). LiDAR data are particularly suited to studying forest structure and dynamics, and has been used to identify patterns in biomass (Asner et al., 2009, Clark et al., 2011, Drake et al., 2002), ecosystem development (Kellner et al., 2011), and successional dynamics (Castillo et al., 2012, Dubayah et al., 2010) in tropical forests. Most promisingly, LiDAR and imaging spectroscopy data have been fused to yield improved estimates of species richness, species distributions, and forest biomass in savanna and forest ecosystems (Asner et al., 2008, Cho et al., 2012, Colgan et al., 2012, Lucas et al., 2008).
We have recently shown that medium-resolution data from Landsat and the Shuttle Radar Topography Mission (SRTM) can be used to infer floristic discontinuities in northwestern Amazonia, corresponding to geological formations and their edaphic properties (Higgins et al., 2011, Higgins et al., 2012). Similar relationships between soils and forest composition in Central America and Asia also suggest that these patterns might be widespread (Jones et al., 2013, Palmiotto et al., 2004). Due to the limited spatial and spectral resolution of Landsat and SRTM data, however, we have been unable to determine whether these compositional discontinuities are translated into patterns in forest structure and functional properties.
Here we propose that imaging spectroscopy and LiDAR may be used both to identify changes in plant species composition, and to provide detailed information about the functional and structural differences between compositionally-defined forest types. To evaluate this possibility, we combined field inventories of plant species composition from a tropical forest in central Panama with spectral and LiDAR data from the Carnegie Airborne Observatory Airborne Taxonomic Mapping System (CAO-AToMS) (Asner, Knapp, et al., 2012). CAO-AToMS incorporates a dual-channel waveform LiDAR with a 428-channel Visible to Shortwave Infrared (VSWIR) imaging spectrometer, providing orthorectified 1 m and 2 m resolution data, respectively, when flown at 2000 m above ground level (a.g.l.).
We concentrated our study on a boundary between two geological formations. Our objectives were (1) to test the ability of imaging spectroscopy and LiDAR to detect changes in plant species composition between these formations; (2) to compare these formations on the basis of forest canopy structure and spectra as identified from LiDAR and VSWIR data; and (3) to examine the relationship between geology or geomorphology and patterns in canopy reflectance and structure. To our knowledge, this is the first time that imaging spectroscopy or LiDAR data have been tested for its ability to detect and study compositionally-defined types in tropical forests.
Section snippets
Study area
We focused our data collection on a boundary between the Ocú and Gatuncillo geological formations in central Panama (Fig. 1), at a site characterized by unbroken, tropical broadleaf rainforest. The Ocú Formation dates approximately to the Paleocene and consists of uplifted limestone and volcanic tuff extending north and east into the Panamanian cordillera (MICI, 1990). The younger Gatuncillo Formation dates to the middle to late Eocene, and consists of younger sedimentary material deposited at
Patterns in CAO-AToMS data
We observed clear discontinuities in the VSWIR and LiDAR data, corresponding to the boundary between the Ocú and Gatuncillo Formations (Fig. 2, Fig. 3, Fig. 4). Using VSWIR bands 830 nm, 1650 nm, and 2220 nm set to red, green, and blue, the Gatuncillo Formation was characterized by blue tones and the Ocú Formation by red tones (Fig. 2a). These differences were pronounced after a principal component transformation of the VSWIR data (Fig. 2b). Setting PCA axes one, two, and three to red, green, and
Discussion
Using 16 sites in central Panama, we found that imaging spectroscopy and LiDAR data from the CAO-AToMS system were able to detect turnover in plant species composition in otherwise uniform closed-canopy and tall tropical forest. Variations in VSWIR imagery predicted up to 67% of the variation in tree species composition in the study area, as measured by a single NMDS axis, and this rose to 76% after the removal of an outlying site. Variations in LiDAR data, furthermore, predicted up to 80% of
Acknowledgments
Data collection and analysis were supported by the Grantham Foundation for the Protection of the Environment and the endowment of the Carnegie Institution for Science. The Carnegie Airborne Observatory is made possible by the Gordon and Betty Moore Foundation; the Grantham Foundation for the Protection of the Environment; the John D. and Catherine T. MacArthur Foundation; the Avatar Alliance Foundation; the W. M. Keck Foundation; the Margaret A. Cargill Foundation; Mary Anne Nyburg Baker and G.
References (49)
- et al.
Carnegie Airborne Observatory-2: Increasing science data dimensionality via high-fidelity multi-sensor fusion
Remote Sensing of Environment
(2012) - et al.
Spectroscopy of canopy chemicals in humid tropical forests
Remote Sensing of Environment
(2011) - et al.
LIDAR remote sensing for secondary tropical dry forest identification
Remote Sensing of Environment
(2012) - et al.
Mapping tree species composition in South African savannas using an integrated airborne spectral and LiDAR system
Remote Sensing of Environment
(2012) - et al.
Hyperspectral discrimination of tropical rain forest tree species at leaf to crown scales
Remote Sensing of Environment
(2005) - et al.
Estimation of tropical rain forest aboveground biomass with small-footprint lidar and hyperspectral sensors
Remote Sensing of Environment
(2011) - et al.
Sensitivity of large-footprint lidar to canopy structure and biomass in a neotropical rainforest
Remote Sensing of Environment
(2002) - et al.
Ecological fingerprinting of ecosystem succession: Estimating secondary tropical dry forest structure and diversity using imaging spectroscopy
Remote Sensing of Environment
(2007) - et al.
Characterizing canopy biochemistry from imaging spectroscopy and its application to ecosystem studies
Remote Sensing of Environment
(2009) - et al.
Classification of Amazonian primary rain forest vegetation using Landsat ETM plus satellite imagery
Remote Sensing of Environment
(2005)
Differences in leaf traits, leaf internal structure, and spectral reflectance between two communities of lianas and trees: Implications for remote sensing in tropical environments
Remote Sensing of Environment
Mapping of continuous floristic gradients in grasslands using hyperspectral imagery
Remote Sensing of Environment
Retrieval of foliar information about plant pigment systems from high resolution spectroscopy
Remote Sensing of Environment
High-resolution mapping of forest carbon stocks in the Colombian Amazon
Biogeosciences
Environmental and biotic controls over aboveground biomass throughout a tropical rain forest
Ecosystems
Invasive plants transform the three-dimensional structure of rain forests
Proceedings of the National Academy of Sciences of the United States of America
Forest canopy gap distributions in the Southern Peruvian Amazon
PLoS ONE
Carnegie Airborne Observatory: In-flight fusion of hyperspectral imaging and waveform light detection and ranging (wLiDAR) for three-dimensional studies of ecosystems
Journal of Applied Remote Sensing
Canopy phylogenetic, chemical and spectral assembly in a lowland Amazonian forest
New Phytologist
A universal airborne LiDAR approach for tropical forest carbon mapping
Oecologia
High-resolution forest carbon stocks and emissions in the Amazon
Proceedings of the National Academy of Sciences of the United States of America
Remote analysis of biological invasion and biogeochemical change
Proceedings of the National Academy of Sciences of the United States of America
Estimating vegetation beta diversity from airborne imaging spectroscopy and unsupervised clustering
Remote Sensing
Hyperspectral remote sensing of canopy biodiversity in Hawaiian lowland rainforests
Ecosystems
Cited by (19)
Mapping the relative abundance of soil microbiome biodiversity from eDNA and remote sensing
2022, Science of Remote SensingCharacterization of indicator tree species in neotropical environments and implications for geological mapping
2018, Remote Sensing of EnvironmentCitation Excerpt :Vitousek et al. (2009) correlated variations in canopy nitrogen content and vegetation height to substrate age using airborne VSWIR hyperspectral and Light Detection And Ranging (LIDAR) data. Higgins et al. (2014b) observed clear relationships between vegetation species composition, VSWIR spectral response and LIDAR structural data on two different underlying sedimentary formations in a Central Panama broadleaf forest. However, the information that can be obtained through hyperspectral remote sensing goes beyond what has been shown so far in GbSR approaches at the Neotropical zone, which hosts hyper-diverse vegetation types.
Forest species diversity mapping using airborne LiDAR and hyperspectral data in a subtropical forest in China
2018, Remote Sensing of EnvironmentCitation Excerpt :However, the low spectral resolution of multispectral sensors cannot adequately capture biophysical and biochemical information of forest canopy species (Carlson et al., 2007). Imaging spectroscopy and LiDAR (Light Detection and Ranging) have become powerful tools for predicting species richness, and even allow for the identification of individual plant species, in different ecosystems (Ceballos et al., 2015; Ghosh et al., 2014; Higgins et al., 2014; Leutner et al., 2012; Lopatin et al., 2015; Vaglio Laurin et al., 2014). Imaging spectroscopy, also called hyperspectral remote sensing, enables continuous sampling of the reflected electromagnetic spectrum from the visual to the shortwave infrared region (Ghosh et al., 2014).
Multi-phenology WorldView-2 imagery improves remote sensing of savannah tree species
2017, International Journal of Applied Earth Observation and GeoinformationCitation Excerpt :Although they can image a very high number of bands, most bands are highly correlated and a few information-rich spectral regions, linked to leaf pigment and other biochemical properties, are more relevant for species discrimination (Sobhan, 2007). High spatial resolution hyperspectral systems have been reported to be an efficient approach for tree species mapping in a variety of biomes or ecosystems (Asner et al., 2008; Oldeland et al., 2010; Naidoo et al., 2012; Higgins et al., 2014). However, airborne hyperspectral tools are not widely available, especially for inventories and management applications.
The spectral variability hypothesis does not hold across landscapes
2017, Remote Sensing of EnvironmentSpectranomics: Emerging science and conservation opportunities at the interface of biodiversity and remote sensing
2016, Global Ecology and ConservationCitation Excerpt :In the first ten years of the Spectranomics project, we have collected, cataloged and stored more than 13,000 canopy tree and liana specimens, in over 3,000,000 tissue samples, representing about 10,000 species heavily biased to tropical sources (Fig. 2). For perspective, this number approaches the total number of tree species in the Amazon basin of roughly 11,000 (Hubbell et al., 2008), a value that would put the global tropical tree inventory at 30,000 species if we liberally extrapolate to the entire Neotropics plus the African and Asian-Oceanic tropics. The Spectranomics database focuses only on those species found in the canopy, meaning they are in full sunlight and are observable from above.