Jean-Baptiste Féret

Tree Species Discrimination in Tropical Forests Using Airborne Imaging Spectroscopy

We identify canopy species in a Hawaiian tropical forest using supervised classification applied to airborne hyperspectral imagery acquired with the Carnegie Airborne Observatory-Alpha system. Nonparametric methods (linear and radial basis function support vector machine, artificial neural network, and k-nearest neighbor) and parametric methods (linear, quadratic, and regularized discriminant analysis) are compared for a range of species richness values and training sample sizes. We find a clear advantage in using regularized discriminant analysis, linear discriminant analysis, and support vector machines. No unique optimal classifier was found for all conditions tested, but we highlight the possibility of improving support vector machine classification with a better optimization of its free parameters. We also confirm that a combination of spectral and spatial information increases accuracy of species classification: we combine segmentation and species classification from regularized discriminant analysis to produce a map of the 17 discriminated species. Finally, we compare different methods to assess spectral separability and find a better ability of Bhattacharyya distance to assess separability within and among species. The results indicate that species mapping is tractable in tropical forests when using high-fidelity imaging spectroscopy.

Mapping Savanna Tree Species at Ecosystem Scales Using Support Vector Machine Classification and BRDF Correction on Airborne Hyperspectral and LiDAR Data

Mapping the spatial distribution of plant species in savannas provides insight into the roles of competition, fire, herbivory, soils and climate in maintaining the biodiversity of these ecosystems. This study focuses on the challenges facing large-scale species mapping using a fusion of Light Detection and Ranging (LiDAR) and hyperspectral imagery. Here we build upon previous work on airborne species detection by using a two-stage support vector machine (SVM) classifier to first predict species from hyperspectral data at the pixel scale. Tree crowns are segmented from the lidar imagery such that crown-level information, such as maximum tree height, can then be combined with the pixel-level species probabilities to predict the species of each tree. An overall prediction accuracy of 76% was achieved for 15 species. We also show that bidirectional reflectance distribution (BRDF) effects caused by anisotropic scattering properties of savanna vegetation can result in flight line artifacts evident in species probability maps, yet these can be largely mitigated by applying a semi-empirical BRDF model to the hyperspectral data. We find that confronting these three challenges—reflectance anisotropy, integration of pixel- and crown-level data, and crown delineation over large areas—enables species mapping at ecosystem scales for monitoring biodiversity and ecosystem function.

Discrete Anisotropic Radiative Transfer (DART 5) for modeling airborne and satellite spectroradiometer and LIDAR acquisitions of natural and urban landscapes

Satellite and airborne optical sensors are increasingly used by scientists, and policy makers, and managers for studying and managing forests, agriculture crops, and urban areas. Their data acquired with given instrumental specifications (spectral resolution, viewing direction, sensor field-of-view, etc.) and for a specific experimental configuration (surface and atmosphere conditions, sun direction, etc.) are commonly translated into qualitative and quantitative Earth surface parameters. However, atmosphere properties and Earth surface 3D architecture often confound their interpretation. Radiative transfer models capable of simulating the Earth and atmosphere complexity are, therefore, ideal tools for linking remotely sensed data to the surface parameters. Still, many existing models are oversimplifying the Earth-atmosphere system interactions and their parameterization of sensor specifications is often neglected or poorly considered. The Discrete Anisotropic Radiative Transfer (DART) model is one of the most comprehensive physically based 3D models simulating the Earth-atmosphere radiation interaction from visible to thermal infrared wavelengths. It has been developed since 1992. It models optical signals at the entrance of imaging radiometers and laser scanners on board of satellites and airplanes, as well as the 3D radiative budget, of urban and natural landscapes for any experimental configuration and instrumental specification. It is freely distributed for research and teaching activities. This paper presents DART physical bases and its latest functionality for simulating imaging spectroscopy of natural and urban landscapes with atmosphere, including the perspective projection of airborne acquisitions and LIght Detection And Ranging (LIDAR) waveform and photon counting signals.

Mapping tropical forest canopy diversity using high‐fidelity imaging spectroscopy

There is a growing need for operational biodiversity mapping methods to quantify and to assess the impact of climate change, habitat alteration, and human activity on ecosystem composition and function. Here, we present an original method for the estimation of α- and β-diversity of tropical forests based on high-fidelity imaging spectroscopy. We acquired imagery over high-diversity Amazonian tropical forest landscapes in Peru with the Carnegie Airborne Observatory and developed an unsupervised method to estimate the Shannon index (H′) and variations in species composition using Bray-Curtis dissimilarity (BC) and nonmetric multidimensional scaling (NMDS). An extensive field plot network was used for the validation of remotely sensed α- and β-diversity. Airborne maps of H′ were highly correlated with field α-diversity estimates (r = 0.86), and BC was estimated with demonstrable accuracy (r = 0.61–0.76). Our findings are the first direct and spatially explicit remotely sensed estimates of α- and β-diversity of humid tropical forests, paving the way for new applications using airborne and space-based imaging spectroscopy.

Landscape‐scale variation in plant community composition of an African savanna from airborne species mapping

Information on landscape-scale patterns in species distributions and community types is vital for ecological science and effective conservation assessment and planning. However, detailed maps of plant community structure at landscape scales seldom exist due to the inability of field-based inventories to map a sufficient number of individuals over large areas. The Carnegie Airborne Observatory (CAO) collected hyperspectral and lidar data over Kruger National Park, South Africa, and these data were used to remotely identify > 500 000 tree and shrub crowns over a 144-km2 landscape using stacked support vector machines. Maps of community compositional variation were produced by ordination and clustering, and the importance of hillslope-scale topo-edaphic variation in shaping community structure was evaluated with redundancy analysis. This remote species identification approach revealed spatially complex patterns in woody plant communities throughout the landscape that could not be directly observed using field-based methods alone. We estimated that topo-edaphic variables representing catenal sequences explained 21% of species compositional variation, while we also uncovered important community patterns that were unrelated to catenas, indicating a large role for other soil-related factors in shaping the savanna community. Our results demonstrate the ability of airborne species identification techniques to map biodiversity for the evaluation of ecological controls on community composition over large landscapes.

Semi-Supervised Methods to Identify Individual Crowns of Lowland Tropical Canopy Species Using Imaging Spectroscopy and LiDAR

Our objective is to identify and map individuals of nine tree species in a Hawaiian lowland tropical forest by comparing the performance of a variety of semi-supervised classifiers. A method was adapted to process hyperspectral imagery, LiDAR intensity variables, and LiDAR-derived canopy height and use them to assess the identification accuracy. We found that semi-supervised Support Vector Machine classification using tensor summation kernel was superior to supervised classification, with demonstrable accuracy for at least eight out of nine species, and for all combinations of data types tested. We also found that the combination of hyperspectral imagery and LiDAR data usually improved species classification. Both LiDAR intensity and LiDAR canopy height proved useful for classification of certain species, but the improvements varied depending upon the species in question. Our results pave the way for target-species identification in tropical forests and other ecosystems.

Predicting leaf gravimetric water content from foliar reflectance across a range of plant species using continuous wavelet analysis

Leaf water content is an important variable for understanding plant physiological properties. This study evaluates a spectral analysis approach, continuous wavelet analysis (CWA), for the spectroscopic estimation of leaf gravimetric water content (GWC, %) and determines robust spectral indicators of GWC across a wide range of plant species from different ecosystems. CWA is both applied to the Leaf Optical Properties Experiment (LOPEX) data set and a synthetic data set consisting of leaf reflectance spectra simulated using the leaf optical properties spectra (PROSPECT) model. The results for the two data sets, including wavelet feature selection and GWC prediction derived using those features, are compared to the results obtained from a previous study for leaf samples collected in the Republic of Panamá (PANAMA), to assess the predictive capabilities and robustness of CWA across species. Furthermore, predictive models of GWC using wavelet features derived from PROSPECT simulations are examined to assess their applicability to measured data. The two measured data sets (LOPEX and PANAMA) reveal five common wavelet feature regions that correlate well with leaf GWC. All three data sets display common wavelet features in three wavelength regions that span 1732-1736 nm at scale 4, 1874-1878 nm at scale 6, and 1338-1341 nm at scale 7 and produce accurate estimates of leaf GWC. This confirms the applicability of the wavelet-based methodology for estimating leaf GWC for leaves representative of various ecosystems. The PROSPECT-derived predictive models perform well on the LOPEX data set but are less successful on the PANAMA data set. The selection of high-scale and low-scale features emphasizes significant changes in both overall amplitude over broad spectral regions and local spectral shape over narrower regions in response to changes in leaf GWC. The wavelet-based spectral analysis tool adds a new dimension to the modeling of plant physiological properties with spectroscopy data.

Detection of Flavescence dorée Grapevine Disease Using Unmanned Aerial Vehicle (UAV) Multispectral Imagery

Flavescence doree is a grapevine disease affecting European vineyards which has severe economic consequences and containing its spread is therefore considered as a major challenge for viticulture. Flavescence doree is subject to mandatory pest control including removal of the infected vines and, in this context, automatic detection of Flavescence doree symptomatic vines by unmanned aerial vehicle (UAV) remote sensing could constitute a key diagnosis instrument for growers. The objective of this paper is to evaluate the feasibility of discriminating the Flavescence doree symptoms in red and white cultivars from healthy vine vegetation using UAV multispectral imagery. Exhaustive ground truth data and UAV multispectral imagery (visible and near-infrared domain) have been acquired in September 2015 over four selected vineyards in Southwest France. Spectral signatures of healthy and symptomatic plants were studied with a set of 20 variables computed from the UAV images (spectral bands, vegetation indices and biophysical parameters) using univariate and multivariate classification approaches. Best results were achieved with red cultivars (both using univariate and multivariate approaches). For white cultivars, results were not satisfactory either for the univariate or the multivariate. Nevertheless, external accuracy assessment show that despite problems of Flavescence doree and healthy pixel misclassification, an operational Flavescence doree mapping technique using UAV-based imagery can still be proposed.

Microtopographic controls on lowland Amazonian canopy diversity from imaging spectroscopy

Microtopographic variation is ubiquitous throughout lowland Amazonia, and it may impart patterns of species richness and abundance, and perhaps community compositional changes. To date, no studies have determined the degree to which lowland microtopography influences forest canopy diversity. We developed the first high-resolution maps of forest canopy diversity in Amazonia, focusing on four landscapes on two river systems in Peru. Spectroscopic images were acquired using the Carnegie Airborne Observatory combined with a new method based on spectral species to map α- and β-diversity. We analyzed spatial patterns in diversity with respect to floodplain and terrace (terra firme) surfaces and in upriver and downriver locations with contrasting landscape morphologies. We found slightly lower average α-diversity in floodplains, but with greater variance than in terrace communities caused by the floodplain mix of swamp forests, anoxic low-diversity ecosystems, and high-diversity areas. β-diversity estimated with the Bray-Curtis dissimilarity (BC) was strongly related to microtopography, with floodplains showing higher internal compositional dissimilarity than terraces. Throughout all landscapes, remotely mapped BC within terrace environments ranged from 0.25 to 0.43, but these values increased 30–77% on floodplains. Upriver landscapes characterized by higher terraces showed more distinct community turnover than did their downstream counterparts. We conclude that microtopography strongly influences β-diversity throughout the study landscapes, but terrain is weakly associated with variation in α-diversity. We uncover the importance of microtopography in determining species composition in lowland Amazonia and highlight the value of imaging spectroscopy for biodiversity research and conservation.

Binary partition tree as a hyperspectral segmentation tool for tropical rainforests

Individual tree crown delineation in tropical forests is of great interest for ecological applications. In this paper we propose a method for hyperspectral image segmentation based on binary tree partitioning. The initial partition is obtained from a watershed transformation in order to make the method computationally more efficient. Then we use a non-parametric region model based on histograms to characterize the regions and the diffusion distance to define the region merging order. The pruning strategy is based on the discontinuity of size increment observed when iteratively merging the regions. The segmentation quality is assessed visually and appears to perform well on most cases, but tree delineation could be improved by including structural information derived from LiDAR data.