Applications in remote sensing—natural landscapes

Applications in remote sensing—natural landscapes

Chapter 3.1 Applications in remote sensingdnatural landscapes Touria Bajjouka, Florian de Boissieub, Jocelyn Chanussotc, Sylvain Doutd, e, Marie Dumo...

3MB Sizes 2 Downloads 113 Views

Chapter 3.1

Applications in remote sensingdnatural landscapes Touria Bajjouka, Florian de Boissieub, Jocelyn Chanussotc, Sylvain Doutd, e, Marie Dumonte, Jean-Baptiste Fe´retb, The´o Massonc, Audrey Minghellif, Pascal Mouquetg, Fre´de´ric Schmidth and Mauro Dalla Murac, i, * a

Laboratoire d’Ecologie Benthique Coˆtie`re (PDG-ODE-DYNECO-LEBCO), Brest, France; TETIS, Irstea, AgroParisTech, CIRAD, CNRS, Universite´ Montpellier, Montpellier, France; cUniv. Grenoble Alpes, CNRS, Grenoble INP, (Institute of Engineering Univ. Grenoble Alpes), GIPSA-Lab, Grenoble, France; dInstitut de Plane´tologie et d’Astrophysique de Grenoble (IPAG), Grenoble, France; eMe´te´o-France-CNRS, CNRM/CEN, Saint Martin d’He`res, France; f Laboratoire des Sciences de l’Information et des Syste`mes (LSIS), University of South Toulon Var ISITV, La Valette, France; gSaint Leu, La Re´union, France; hGEOPS, Univ. Paris-Sud, CNRS, Universite´ Paris-Saclay, Orsay, France; iTokyo Tech World Research Hub Initiative (WRHI), School of Computing, Tokyo Institute of Technology, Tokyo, Japan *Corresponding author. b

1. Introduction Hyperspectral imagery is a booming technology with a great impact in many different thematic fields. One of the main domains in which this acquisition technology finds a very large spread is remote sensing [1,2]. The high spectral resolution typical of hyperspectral acquisitions allows one to access spectral properties (e.g., reflectance) of objects in a scene of interests. Accessing such information is a proxy for gathering some information of physical and chemical properties of materials becoming of utmost importance in many applications in Earth and planetary observations. For example, applications such as land cover mapping, detection of target or anomalous materials, estimation of higher level geophysical variables (e.g., physicochemical properties of materials, morphological features, and biodiversity) can be carried out based on such input information. In this scenario, hyperspectral imagery becomes a very relevant tool that can aid practitioners and end users in gathering information on a natural or anthropic phenomenon of interest. Hyperspectral Imaging. https://doi.org/10.1016/B978-0-444-63977-6.00016-X Copyright © 2020 Elsevier B.V. All rights reserved.

371

372 SECTION j III Application fields

Although an exhaustive overview of all applications of hyperspectral imaging (HSI) in remote sensing is not possible, we aim at giving the reader a glance of some selected common research fields in which this technology is used, contextualizing each application with a brief overview of the context, presenting an example of real case study and some discussion. The presentation of the use of hyperspectral imagery in remote sensing applications spans two chapters. Applications are grouped in the monitoring of natural resources and anthropic activities. This chapter will present applications related to the observation of natural scenes aiming at the monitoring of natural resources or inspection of scenes. Four applications are covered in this chapter: 1. 2. 3. 4.

Planetary surfaces Coastal areas Cryosphere Vegetation

2. Planetary sciences 2.1 Context Since the 1980s, visible and infrared HSI is a privileged technique for exploring planetary surfaces and atmospheres. The spectral capabilities allow to identify chemical components, even the ones that were not expected before launch. They form materials or aerosols whose physical state, particle size distribution, and stratification can also be constrained. The imaging capabilities provide the spatial extent of the components and of their properties revealing terrain units. For decades, the spectral/spatial resolution and sampling have been improved significantly. The detection of species related to small geological objects (faults, rocky outcrops, and dunes) becomes possible. Currently the capabilities of hyperspectral sensors are limited by data downlink constraints, especially when exploring the external part of the Solar System. The list of the main remote sensing hyperspectral instruments that have been operated is available in Table 1. The spectra acquired in the near-infrared by VIRTIS above Venus are dominated by the atmosphere contribution. They contain information to study the dynamics and the chemistry of the different atmospheric layers [19,20]. The pioneer near-infrared hyperspectral images of Mars allowed to map surface features, such as hydration of the soil [3]. Indirect measurement of altitude was also possible, thanks to local pressure estimation. Indeed atmospheric CO2 gas leaves on the spectra absorption bands that are sensitive to the local surface pressure. The next generation of instruments (OMEGA and CRISM) revealed more diversity at the surface of the Red Planet [21,22]. Most studies focused on detection at the surface of various families of minerals such as

TABLE 1 List of the main remote sensing imaging spectrometer instruments (more than hundreds of spectral channels) in Earth and Planetary science, with their main characteristics (date of arrival, mission platform, planetary body, spectral range, maximum number of recorded wavelength). Date

Mission

Body

Wavelength (mm)

Max number channels

ISM [3]

1989

Phobos 2

Mars

0.76e3.14

128

TES [4]

1992

Mars Global surveyor

Mars

5.8e150

286

NIMS [5]

1992

Galileo

Jupiter

0.7e5.2

408

AVIRIS [6]

1993

Airborne

Earth

0.400e2.500

224

VIRTIS [7]

1998

Rosetta

67P/ChuryumovGerasimenko

0.25e5.0

388

Hyperion [8]

2003

Earth Observing 1

Earth

0.400e2.500

242

OMEGA [9]

2004

Mars Express

Mars

0.38e5.1

352

VIMS [10]

2004

Cassini

Saturn

0.3e5.1

352

CRISM [11]

2007

Mars Reconnaissance Orbiter

Mars

0.362e3.920

544

VIRTIS [12]

2007

Venus Express

Venus

0.28e5.0

432

M [13]

2009

Chandrayaan-1

Moon

0.430e3.000

260

VIR [14]

2011

DAWN

Ceres

0.25e5.0

864

RALPH/LEISA [15]

2015

New Horizons

Pluton/Triton

1.25e2.25

256

3

373

Continued

Applications in remote sensingdnatural landscapes Chapter j 3.1

Name

Name

Date

Mission

Body

Wavelength (mm)

Max number channels

SIMBIO-SYS [16]

2025

BepiColombo

Mercury

0.4e2.0

256

MERTIS [17]

2025

BepiColombo

Mercury

7e14

78

MAJIS [18]

2030

JUICE

Jupiter

0.4e5.7

w1200

Other characteristics are available in the corresponding bibliographic references.

374 SECTION j III Application fields

TABLE 1 List of the main remote sensing imaging spectrometer instruments (more than hundreds of spectral channels) in Earth and Planetary science, with their main characteristics (date of arrival, mission platform, planetary body, spectral range, maximum number of recorded wavelength).dcont’d

Applications in remote sensingdnatural landscapes Chapter j 3.1

375

mafics, clays, sulfates, salts, as well as water and carbon dioxide ices. The spectra also contain information on their abundances and microtexture (type of mixture, grain size, surface roughness,etc.). For instance, the North polar cap of Mars, which is mainly composed of H2O ice, shows an evolution of grain size due to solar irradiation and surface/atmosphere exchanges in summer [23]. The OMEGA instrument also performed the first unambiguous observations of CO2 ice clouds on Mars [24]. New insights about the nature and activity of the satellites of Jupiter came from the images of the Near-Infrared Mapping Spectrometer (NIMS) aboard Galileo. The instrument revealed that Io is a body continuously covered by hundreds of hot spots linked to the functioning of volcanoes. The eruption styles of the most remarkable individuals were constrained, thanks to the modeling of the emission component of the spectra [25]. Reflectance spectra showed that Io’s crust is dominated by sulfur-bearing species, in particular SO2 frost [26]. Europa’s crust is essentially composed of water ice mixed in various proportions with sulfur compounds, hydrated sulfuric acid, hydrated sulfate, and chloride salts [27]. The composition reflects the spatially varying influences of the internal geologic activity and exogenic processes such as the bombardment of the surface by Jupiter’s magnetospheric particles. The long-lived Cassini mission to Saturn’s system carried the imaging spectrometer VIMS that acquired a large collection of images of the icy moons with particular emphasis on the main representative Titan. With atmospheric pressures of 1.5 bar and temperatures of 90e95 K at the surface, methane and ethane condense out of Titan’s nitrogendominated atmosphere and flow as liquids on the surface. VIMS was a key instrument to monitor the global spatial coverage of the CH4 and C2H6 clouds during half of a seasonal cycle [28]. Along with the other imagers of Cassini, it also shaded light on the fascinating terrains of Titan such as vast equatorial dunes, plateaus incised by well-organized channel networks, lakes and seas of liquid hydrocarbons, all related to climatic and geologic processes [29]. To conclude this quick survey, one should also mention the recent breakthroughs accomplished in the far reaches of the Solar System during the flyby of Pluto, its main satellite Charon, and four other companions by the space probe New Horizons in July 2015. Pluto’s surface was found to display a wide variety of landforms and terrain ages, as well as dramatic albedo, color, and compositional variegation revealed by Ralph the infrared spectro-imager [30]. The previous discoveries have been sustained by a wide range of methodologies for analyzing the data. Apart from the simple band ratio techniques, other approach using wavelet transform has been developed [31]. Also, supervised linear unmixing methods have permitted better surface characterization, incorporating mathematical constraints such as positivity [32,33]. Despite the nonlinear nature of the radiative transfer, some arrangement has been proposed to minimize the nonlinearities and allows robust mapping in the linear approximation [34]. In parallel, blind separation techniques were developed since the set of reference spectra from the lab may be not relevant in planetary

376 SECTION j III Application fields

exploration. The principal component analysis (PCA) was used for unsupervised analysis of the data set [35]. Nevertheless, since spectra endmembers and abundances maps are correlated, PCA is not the most relevant tool. Independent component analysis was used to relax the orthogonality and helped to decipher Venus’ signal [36]. The most comprehensive step was performed by incorporating the positivity and sum-to-one constraints. A pioneer work demonstrated that such algorithms are able to estimate automatically maps and endmembers spectra that were consistent with laboratory measurements [37]. Recent advances in instrumentation allow acquisition of hyperspectral images at different angular geometries over the same site. Thanks to these capabilities, the CRISM instrument permitted the estimation of the optical constant of the dust aerosols on Mars and the estimation of the total dust content of the atmosphere [38]. After a correction of the atmospheric effect, the mapping of the surface bidirectional reflectance distribution function was possible [39,40], allowing to map the surface microtexture by remote sensing [41].

2.2 An example of application: mapping Mars surface On Mars, CO2 is a major component of the atmosphere whether in the gaseous state (abundance of 95%) or in the solid state at the surface. During winter, up to 30% of the atmospheric CO2 condenses or precipitates to form seasonal polar deposits. They completely sublimate in spring except in the South pole where a residual cap persists throughout the summer. Recent studies have shown that these condensates, often in the form of transparent compact layers, cause during their sublimation the erosion of their substratum, the functioning of dusty plumes, and the triggering of avalanches on the slopes. Thus, the Martian CO2 is suspected of being a significant geological agent at present. A representative case study is the megadune located in the Russell Crater of Mars. This place is the siege of a complex defrosting sequence each spring implying seasonal CO2 and a small amount of water ice [42]. Besides on its pole facing slope, the dune displays gullies thought to have been carved by liquid water although they could also be related to dry avalanches triggered by the defrosting activity. In order to improve our understanding of these phenomena, hyperspectral and high-resolution panchromatic imaging are used to monitor the dune during spring. The hyperspectral sensor is the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) and the panchromatic sensor is the High Resolution Imaging Science Experiment (HiRISE), both aboard the Mars Reconnaissance Orbiter of NASA. In the targeted mode, CRISM senses planet Mars at 18 m pixel1 using 437 visible and infrared wavelengths. A key preliminary step is the fine coregistration of both types of imagery. Automatic extraction and characterization of sublimation structures (e.g., dark spots) in the HiRISE images is performed with a multiscale method in several stages: pixel shape index classification [43], geometric filtering, and regularization. The CO2 ice is mapped by applying the unsupervised linear spectral unmixing (SU) methods VCA and MVC-NMF [44] on the CRISM

Applications in remote sensingdnatural landscapes Chapter j 3.1

377

images. The endmembers are spectra representative of different physical conditions of the ice deconvolved of effects related to the limited spatial resolution of the CRISM sensor. The spatial distribution of the endmembers is computed in the form of maps giving their respective surface fraction in each pixel. The physical characterization of the ice comes from the modeling of the endmember spectra with a reflectance physical model [45e47]. The temporal dimension of observations allows us to locate and quantify the changes. Understanding of the defrosting phenomena comes from the joint interpretation of the different products derived from the analysis of the CRISM and HiRISE data: (1) the dark spot distribution maps derived from the morphological analysis; (2) the corresponding estimation of the growth (shrinking) rate of the spots as a function of the position on the dune; (3) the spatial distribution of the different physical conditions recognized by SU; (4) the structural organization of the ice and its level of dust contamination for each ice conditions. Fig. 1 superimposes two products characterizing at a given date on the Russell’s dune: (1) the detection mask of active areadin primary bluedidentified by the morphological analysis on HiRISE highresolution image; (2) colored composition from the SU of the CRISM hyperspectral image indicating the spatial distribution of different physical conditions of CO2 ice and its degree of contamination by dust. We note a very intense activity close to the crest of the dune with the growing of numerous individual spots (primary blue). Conjointly the abundance maps show a high superficial contamination of the ice by the dust (pinkish area). On the main slope, numerous small spots are appearing where granular ice is abundant. We also note the presence of a more compact form of ice in some places (greenish area). On the main slope, flow-like features underlined by reddish hues indicative of very high level of dust contamination are associated with the deepest gullies. However they do

FIGURE 1 Three-dimensional oblique view of a portion of the Russell dune on Mars. Information on the nature of the materials and on their texture as well as on the active areas is extracted from different types of imagery and represented using color coding. Results from X. Ceamanos, S. Doute, B. Luo, F. Schmidt, G. Jouannic, J. Chanussot, Intercomparison and validation of techniques for spectral unmixing of hyperspectral images: a planetary case study, Geoscience and Remote Sensing, IEEE Transactions on 49 (11) (2011) 4341e4358, https://doi.org/10.1109/TGRS. 2011.2140377.

378 SECTION j III Application fields

not show substantial spot growth except in their upper part. Consequently, we can infer that the dust contamination on the slope of the dune has two distinct origins: the activity of the neighboring spots (local source) on the one hand and the avalanches of dust coming from the crest (“distant” source) on the other hand. We also note the presence of limited area with granular weakly contaminated CO2 ice (light blue area). We interpret the latter units as locations where CO2 gas emitted by sublimation in the vicinity is recondensing on colder slopes. Fig. 2 represents the spectral endmembers. CRISM hyperspectral imaging and HiRISE panchromatic high-resolution imaging of the Russell dune on Mars greatly provided details about the evolution of the CO2 ice, physical conditions, and degree of contamination by dust. Thus they document clearly the associated phenomena prior to the complete sublimation of these seasonal deposits. The main conclusion of our study is that intense ejection of dust occurs in multiple places on the crest of the Russell dune because of the CO2 sublimation. This phenomenon leads to the accumulation of a thick dust layer on the ice that causes dust avalanches channeled in the gullies. Thus it is now established that the gullies are currently not active by the flow of water but by the flow of dust. This science case illustrates the decisive contribution of HSI to understand planetary processes. It also demonstrates the interest of conjugating compositional information with textural and geometrical information extracting from panchromatic imaging with higher spatial resolution.

(A)

(B)

(C)

(D)

FIGURE 2 Spectral endmembers extracted from the data set presented in Fig. 1 using different methods (A) VCA (vertex component analysis), (B) BPSS (Bayesian positive source separation), (C) MVC-NMF (minimum volume constrained non-negative matrix factorization), and (D) spatialVCA (vertex component analysis) methods (see Ref. [44] for more details).

Applications in remote sensingdnatural landscapes Chapter j 3.1

379

2.3 Open challenges in planetary imaging spectroscopy The history of planetary exploration clearly shows the benefit of improving the capabilities of the imaging spectrometers. Their spectral range and resolution, swath width, and spatial resolution, signal-to-noise ratio (SNR), and photometric accuracy have gradually improved, mainly at the pace of detector developments. However, the improvement of the spectral resolution has been very limited during the last 30 years (factor <2.5) and is still far from what would be necessary to fully resolve a large number of ice bands (CO2, CH4, SO2, CO, and CxHy) or to allow the identification of minor components, even with future spectro-imagers under construction for missions in the 2020e30 horizon (e.g., JUICE and Mars2020). The second major handicap of the current instruments is their weight, which has been multiplied by a factor of four to six in 20 years, currently reaching 20e30 kg for a complete Vis-NIR instrument with a resolution capping at z300 @ 2 mm. These weights are hardly compatible with lightweight missions such as Mars’ in situ exploration programs (ExoMars program/ESA, March 2020/NASA, Insight/NASA, etc.). It is therefore necessary to find new technologies that can both drastically reduce the weight (but also the volume and consumption) while enabling new science with spectroimagers reaching a resolving power R of z1000 @ 2 mm. As a matter of fact, miniaturized low-cost spectro-imagers could be deployed in large numbers aboard fleet of nanosatellites, allowing new observation strategies like the systematic acquisition of multiangular sequences. Future instrument could also be enriched by the measure of polarization for a better characterization of the aerosol contribution and the microtexture of the surface. These technological breakthroughs would require significant advances from the methodological point of view because of increased volume and complexity of data that could not be fully transmitted to Earth due to bandwidth limitations. This calls for onboard intelligent systems able to perform automatically first analysis aimed at returning only the relevant data. On the other hand, accurate and fast processing of highly dimensional data is an ongoing challenge. Finally, the quantitative analysis of high-resolution images implies the use of physical models treating 3D complex radiative transfer effects that influence the images.

3. Coastal areas 3.1 Context Coastal water ecosystems are subject to high spatial and temporal variability in their bio-optical (e.g., absorption and scattering), morphological, and biogeochemical (e.g., phytoplankton biomass, mineral-like hydrosols) properties. As recognized in the Convention on Biological Diversity (CBD), global biodiversity loss continues to escalate, including in coastal ecosystems. More than 50% of the world’s population lives within a 60 km coastal belt (Fig. 3). The impact of increased urbanization and population near the coast has serious implications

380 SECTION j III Application fields

FIGURE 3 Global distribution of coastal and inland aquatic ecosystems. Red indicates regions where water depth is less than 50 m and where land elevation is less than 50 m. Light violet to dark violet gives the concentration of inland wetlands, lakes, rivers, and other aquatic systems. Increased darkness means greater percentage of areal coverage for inland aquatic ecosystems [48,49].

for the maintenance of the ecological quality of coastal waters. It is therefore of utmost importance to observe these areas at high-spatial resolutions, large coverage, and frequent intervals to better understand the functioning of these complex dynamic ecosystems. Such observations are required inputs for mesoscale physical models to predict the evolution of these ecosystems, which is needed for coastal socioeconomic management based on biological indicators. More importantly, major gaps in knowledge remain on the extent, fragmentation, and degradation of coastal marine habitats, particularly for those that harbor high biodiversity and are vulnerable to damage. Among these includeddespite their worldwide-recognized importance and increased protection measuresdcoral reefs, of which 15% will disappear in the coming decades, and seagrass beds, which show an annual decline of 2%e5% of their area worldwide [50]. Hyperspectral images can provide highly valuable data on coastal areas to help fill these gaps owing to their ability to distinguish species and functional traits in aquatic environments and to assess substrate composition. These important details usually cannot be discerned with multispectral sensors which have a limited number of wavelengths or wide wavebands. Compared with open oceans, coastal waters are generally more complex environments because various parameters influence the remotely sensed reflectance, such as bathymetry, water quality, seabed type, water surface, and atmospheric conditions. In shallow waters, the estimation of each of these parameters is easier using hyperspectral than multispectral images [51]. The higher number of spectral bands as well as the increased spectral resolution reduces the confounding effects between optically active parameters [52e56]. In coastal environments, several studies have demonstrated the ability of hyperspectral tool for characterizing and mapping the coastal substrates and benthic communities. Applications of hyperspectral remote sensing include (1) microphytobenthos detection and biomass quantification, as its high spectral resolution is sensitive to the narrow absorption bands of the microphytobenthos [57e59] and (2) the narrow bandwidths of hyperspectral sensors can detect the local features of kelp and macroalgal stands

Applications in remote sensingdnatural landscapes Chapter j 3.1

381

[60]. Furthermore, hyperspectral data demonstrated its usefulness in discriminating among seagrass species and quantifying not just their cover but also their productivity [61,62]; other studies have provided similar data for the vulnerable coral reef ecosystem as well [54,63,64]. Hyperspectral sensors have also been deployed on underwater vehicles to map deep and dark seafloors where passive remote sensing techniques are inconsistent [65]. In addition to their ability to characterize benthic habitats, hyperspectral remote sensing methods can also retrieve data on bathymetry, water quality, and benthic cover. These methods are usually based on a radiative transfer model that describes how light propagates in water [66], thus constituting an inverse problem to determine the causes of the measured patterns of light propagation. The inverse problem can be solved using either lookup tables (LUTs) or iterative optimization [67]. In the first case, a spectral library is built from different combinations of depth, water quality, and benthic cover as inputs of an exact [66] or an approximated [68] radiative transfer model. For each pixel, the measured reflectance is then associated with the closest simulated spectrum in the LUT. Two examples of such approaches include the comprehensive reflectance inversion based on spectrum matching and table lookup (CRISTAL) [69] and adaptive linearized lookup trees (ALLUT) [70] (see Ref. [67]). Alternatively, the inverse problem can be solved by minimizing a cost function between measured and simulated reflectance spectra. In this case, the forward model used for simulation must be sufficiently fast to permit multiple runs for each pixel. Many analytical and semi-analytical models have been developed under various assumptions and water types [68,71,72]. These models approximate the radiative transfer equation and generally simulate the reflectance of shallow waters as a function of illumination/observation geometry, depth, bottom albedo, and optical properties inherent to the water column (i.e., its absorption and scattering properties). Various cost functions can be used to assess the goodness-of-fit between the observation and the model, including Euclidean distance, spectral angle mapper distance, or maximum likelihood when using LUTs [53,69,70] or iterative optimization to invert the model [67,73e78].

3.2 Example of application: coral reef monitoring case study Shallow coral reefs form some of the most diverse ecosystems on Earth. They occupy less than 0.1% of the world’s ocean surface, but they support at least 25% of all marine species. However, coral reefs are more sensitive than most other coastal environments to threats of anthropogenic origin [79]. They also suffer from high mortality due to coral bleaching in response to increased seawater temperatures [80]. Thus, monitoring these ecosystems is a major issue in the context of global change and increasing ocean acidification. In Reunion Island (Indian Ocean), where the narrow reef has suffered due to human activities, the field-based monitoring conducted on a few stations as a part of the Global Coral Reef Monitoring Network does not take into account the strong spatial heterogeneity that characterizes these reef ecosystems. Hence, there is a

382 SECTION j III Application fields

need to develop a new spatial approach based on imaging. Many in situ or laboratory spectro-radiometric studies have proven able to differentiate between different species, different substrates, and coral health [54,63,81,82]. Thus, owing to their spectral richness and spatial resolution, hyperspectral airborne sensors offer new opportunities to monitor coral reef ecosystems. An example of the spectra of the materials of interest for this analysis is reported in Fig. 4. To characterize and quantify the changes in Reunion Island reef flat habitats and reef features, hyperspectral data have been acquired in situ [83] and over the Reunion Island reef flat through two projects Spectrhabent [83] and Hyscores [84], respectively, in 2009 and 2015. Hyperspectral images were acquired using airborne sensors (Hyspex VNIR-1600 and AISA Eagle 1k systems) with a spatial resolution of 0.4 m and spectral resolution of about 8 nm between 393 and 963 nm. Both data sets were atmospherically corrected and orthorectified before processing. A water column correction, based on a simplified signal attenuation equation derived from Lee’s model [68], associated with a linear unmixing model, were used to extract information related to both depth and abundances of the main seabed types, namely corals, algae, seagrass, and sand. Applied image processing was used to estimate the bathymetry and the coral cover rate, and also to propose a spatial index of coral health called “hyperspectral coral vitality,” expressed as the ratio of coral cover to the sum of coral and algae covers [85]. This approach can take into account the high spatial variability over the entire reef platform. The acquisitions made in 2009 and 2015 were used to assess the changes that took place over a 5-year period. It was thus possible, through a

FIGURE 4 Example of spectra of the main materials of interest in the campaign of acquisition.

Applications in remote sensingdnatural landscapes Chapter j 3.1

383

FIGURE 5 From left to right: coral reef, hyperspectral image acquired over Saint Gilles reef flat (Reunion island, Indian Ocean), bathymetry estimation, classified coral vitality index distribution from low (orange) to high (blue green) values, and evolution of the coral cover with areas of degradation in red and progression in blue.

diachronic analysis, to locate and quantify changes in terms of gains and losses (Fig. 5), both for coral cover and coral vitality as well as for coral geomorphology, opening up new prospects for operational large-scale monitoring. The processing method being initially limited to lagoon reefs, recent work has made significant improvements, based on a semi-analytical model that provides sea surface reflectance as a function of chlorophyll, suspended matter concentrations, colored dissolved organic matter, optical properties of hydrosols, bathymetry, and seabed reflectance [86]. Through an inversion process under the assumption of linear mixing model, it was possible to estimate the bathymetry up to 25 m depth as well as the abundance of the main bottom types up to 10 m depth (Fig. 6). These results show that the accuracy and robustness of the bathymetric estimation were greatly influenced by the chosen inversion process, namely cost functions and physical constraints imposed on the seabed-type abundance retrieval. Nonetheless, the method implemented in this study only differentiated between four benthic types. Future studies may explore the possibility of distinguishing between different coral types (e.g., blue vs. brown, branched vs. massive, hard vs. soft) as well as algal groups.

3.3 Challenges in coastal applications of hyperspectral sensing The first challenge to using HSI is sea surface reflectance. The reflection of the sun rays on the sea surface causes the sea surface reflectance to increase strongly leading to noise in the images. However, several methods using nearinfrared bands can remove this parasitic signal and allow the use of images acquired in nonoptimal conditions (geometric conditions, waves, and wind) [87]. Another challenge in the use of this method is the lack of prior

384 SECTION j III Application fields

FIGURE 6 True color atmospherically corrected hyperspectral images (R ¼ 638 nm, G ¼ 551 nm, B ¼ 471 nm) of (top left) Boucan and (bottom left) Ermitage in Reunion Island, hyperspectral derived bathymetry (to right), and RGB composition of unmixing results for corals, algae, and sand abundances corresponding, respectively, to red, green, and blue channels (bottom right). Seagrass (absent in outer reef area) is indirectly represented by the dark pixels (modified from Petit et al. [86]).

knowledge of the seabed cover. Most methods require prior knowledge on the nature or the spectra of the cover types likely to be found on a specific site. However, this information is not always known for all study sites. Only a few methods consider seabed spectra as an unknown and include this parameter in the inversion model [88]. As for land applications, several coastal application methods use hyperspectral images to estimate the composition of the seabed cover within the pixel considered as a mixture of different pure seabed covers. Most of the time, the mixture is considered as linear [73,74,76,89], summing to unity or not. Nevertheless, the water column is known to have nonlinear effects. Water turbidity increases water attenuation, preventing the signal coming from the bottom from reaching the sea surface. For example, in the Atlantic Ocean, the bathymetry can be estimated to 10 m depth, but estimates can reach 20 m in the Mediterranean Sea or 40 m in oligotrophic seas. The presence of suspended inorganic matter can also hide the presence of chlorophyll and dissolved organic matter. The estimation of the concentration of these elements is thus challenging in turbid waters. Providing reliable uncertainties for remote sensing products is a third important challenge because they determine the confidence level of these products, which is critical for many applications. Uncertainties in estimated ocean color parameters result from the propagation of uncertainties in the bio-optical reflectance modeling through the inversion process. Based on given bio-optical and noise probabilistic models, Crame´reRao bounds can be computed efficiently for any set of ocean color parameters and any sensor configuration, directly providing the minimum estimation variance that can be possibly attained by any unbiased estimator of any targeted parameter [90]. Finally, the majority of hyperspectral sensors are airborne. A high spectral and spatial resolution sensor on a satellite

Applications in remote sensingdnatural landscapes Chapter j 3.1

385

can potentially cover terrestrial and coastal zones. However, for coastal applications, sensors need a high SNR to be able to measure even very weak effects of parameters on surface reflectance. It should be highlighted that the information content in hyperspectral data sets is rich and will present many challenges for operational use by coastal area managers. In addition to the scientific research uses of hyperspectral data, application-specific algorithms can be developed to remove the redundant or undesired spectral information while preserving the desired, targeted information [91].

4. Cryosphere 4.1 Context The characterization of snow extent is critical for a wide range of applications. Thematic fields, such as climatology [92], meteorology [93], energy, human resources monitoring [94], depend by an accurate evaluation of the snow cover (e.g., extent, type of snow, depth of the snow layer, etc.). Here we focus on the estimation of the snow cover extent, which can be relatively large over the Earth and feature large variation in terms of covered surface during the year, especially in the northern hemisphere as shown by Fig. 7. Historically, snow cover monitoring was mostly based on field measurements as data acquired by permanent stations, terrain campaigns, and with aerial acquisitions [95,96]. These data are usually spatially or temporally discontinuous. This is a problem with respect to the high spatial and temporal variability that can experience snow covered. Taking advantage of the peculiar optical properties of snow in comparison with those of other materials, satellites equipped with optical sensors are suitable platforms to monitor snow over large areas and with

FIGURE 7 Typical minimal and maximal snow extent over the northern hemisphere. Source: NASA.

386 SECTION j III Application fields

periodic surveys. In greater details, snow has a unique characteristic in terms of high reflectance in the visible spectrum (Vis, 400e800 nm) and a high adsorbance in the near-infrared (NIR, 800e1000 nm) and short-wave infrared (SWIR, 1000e2500 nm) domains of the electromagnetic spectrum. A large number of satellites are equipped with optical imaging systems that cover the optical and reflective infrared domains (e.g., SPOT 4 and 5, Landsat 8, MODIS on Terra, and Sentinel-2). The data acquired by these satellites are currently used to retrieve the snow cover area (SCA) (i.e., a map showing the snow cover extent) based on the optical properties of snow in the Vis and NIR/SWIR domains. The most usual product at this time comes from the ModerateResolution Imaging Spectroradiometer (MODIS). MODIS offers a ground spatial resolution ranging from 250 m to 1 km at nadir in the Vis and SWIR bands, respectively, and a near-daily return time which increases the probability of cloud-free image. Although MODIS cannot be considered as a true hyperspectral sensor since it does not provide hundreds of spectral bands, it acquires 36 bands (2 at 250 m, 7 at 500 m, and 17 at 1 km) with spectral bandwidths ranging from 50 to 10 nm, which is comparable to the typical bandwidths of hyperspectral cameras (about 10e20 nm). The relatively narrow spectral bands with respect to other multispectral sensors are close to those of hyperspectral acquisitions. The use of MODIS images can show the interest and potential impact in using true hyperspectral data for the analysis. In the rest of this section, we will consider MODIS as it is de facto the optical sensor of reference for estimating snow cover extent. The MODIS sensor onboard of the Terra and Aqua satellites has been in orbit for over 15 years, and many methods have been proposed to produce from MODIS images, a binary SCA map (i.e., indicating the presence/absence of snow in each pixel) or a fractional snow cover fraction (SCF) product (i.e., percentage of snow in each pixel). Salomonsons et al. [97] proposed a method for producing binary snow maps at 500 m. This approach takes advantage of the contrasting reflectance of snow in the SWIR band (MODIS band 6) and green visible band (MODIS band 4) using the Normalized Difference Snow Index (NDSI) [98]. The binary snow cover product is obtained by thresholding the NDSI. Salomonsons et al. [97] retrieved SCF at the pixel level from a linear regression of the NDSI. Landsat images were used both for the calibration of the linear regression and the validation through a binary snow product at 30 m. These aforementioned techniques based on the NDSI for retrieving SCA and SCF maps are currently implemented in the processing chains at NASA for providing snow cover products.1 Despite their large diffusion, approaches of snow cover mapping relying on the NDSI only exploit two of the five available spectral bands at 500 m, not taking full advantage of all the available information in the acquisitions. More recently, several studies

1. https://modis.gsfc.nasa.gov/data/dataprod/mod10.php.

Applications in remote sensingdnatural landscapes Chapter j 3.1

387

started to use the whole spectral range available to produce snow cover maps using SU [99,100]. SU can be formulated as a blind source separation problem with the objective of recovering the spectral signature of some materials, called endmembers, in the scene [101] and their proportions (abundances) in each pixel. In this scenario, some of the endmembers will be associated to snow spectra, leading to an estimation of the SCF in each pixel as the sum of the abundances of the snow endmembers [102]. An SCF product over different places of the globe (e.g., North America, Europe) was developed by Painter et al. [99], who proposed to use SU to produce SCF at 500 m resolution through the so-called MODSCAG processing chain. Pascal et al. [103] used the two MODIS bands at 250 m to improve the spatial resolution of the 500-m bands via image fusion (based on the ARSIS concept [104]), leading to seven bands at 250 m. The authors then proposed the MODIS Imagery Laboratory (MODImLAB) processing chain [100], which is able to produce SCF maps at 250 m through SU using a set of eight spectra obtained from a spectral library as endmembers. In the following, we will present a technique based on SU for the estimation of SCF as an example of application.

4.2 Example of application: snow cover map estimation The area under study is situated in the French Alps near the city of Grenoble. The estimation of SCA maps in this zone is of great importance for meteorological purposes and for the prediction of energy production by hydroelectric plants. This geographical region is characterized by heterogeneous land covers in a mountainous topography with elevations ranging from 200 to 4000m above sea level. An example of image acquired by MODIS on the area of study is reported in Fig. 8. The figure shows the reflectance of the seven bands at 250 m of spatial resolution. The 500-m bands were fused at 250 m following the procedure in Ref. [103]. Snow covered areas can be noticed in correspondence of the mountainous reliefs (mainly on the right part of the image) showing a relatively high reflectance in the visible bands (1e4) and a low reflectance in the near-infrared bands (6e7). In this study, we aim at estimating a snow cover fraction in each pixel using SU. To this end, we consider a set of spectra coming from the spectral library used in Ref. [100] as endmembers (see Fig. 8). When considering an SU-based technique, we could estimate the abundances of the materials from the spectral library following a full least square spectral unmixing (i.e., least square solution of the linear unmixing problem enforcing on the abundance nonnegativity and sum to one on each pixel). Although this approach provides subpixel information on the snow cover, the results might not be totally satisfying. In general, a large number of false detection can be found in areas where there is no snow (e.g., in lower altitude regions). These false detections can be mainly due to the spatial variability of the different materials in the area and to a misfit of the considered linear

388 SECTION j III Application fields

FIGURE 8 (Top) Reflectance of the first seven bands of MODIS over an area of 100 by 80 km near Grenoble, France. Pixels covered by clouds are masked and reported in white. (Bottom) Reflectance of the seven materials considered as endmembers for spectral unmixing. From S. Pascal, R. Mathieu, Y. Arnaud, Subpixel monitoring of the seasonal snow cover with MODIS at 250 m spatial resolution in the southern alps of New Zealand: methodology and accuracy assessment, Remote Sensing of Environment 113 (1) (2009) 160e181.

mixing model with the real data. Furthermore, the relatively low number of spectral bands available from MODIS acquisitions (here seven) limits the capacity to precisely characterize materials in the scene or to use more larger spectral libraries as endmembers. Limitations would be reduced if one would dispose of a true hyperspectral image. In order to prevent false positive detection, a simple solution could be to hard threshold snow abundances which are too low. As these errors are small, in Ref. [99] it was proposed to suppress all the snow fractions below 15% in the SCF map. A less radical approach proposed in Ref. [100] is to threshold snow abundances based on the NDSI value of each pixel. Specifically, if a pixel has an NDSI below 0.2, it is very unlikely that it contains snow and then it will be considered as a snow-free pixel in final SCF map. The result of the approach by Pascal et al. [100] is shown in Fig. 9 where it is possible to appreciate the advantage when using an SU-based approach for producing fractional snow cover maps with respect to techniques producing binary snow cover maps (as by thresholding the NDSI). A critical aspect in this study is the quantitative evaluation of the results. When focusing on the medium spatial resolution of MODIs (250m), a solution

Applications in remote sensingdnatural landscapes Chapter j 3.1

389

FIGURE 9 Results of a binary detection of the snow cover from a simple thresholding of the NDSI (left, snow in yellow), and fractional results obtained by spectral unmixing, following the approach of Pascal et al. [100]. NDSI, normalized difference snow index.

is to use a higher (e.g., decametric) spatial resolution image as those acquired by SPOT 4, SPOT 5, or Landsat 8. These sensors have a low-frequency revisit time (about a week), which is inadequate for following the snow variability, but allows for punctual comparisons. There is consequently a possibility to create binary snow cover maps from decametric spatial resolution images and generate fractional maps once spatially degraded to meet the spatial resolution of MODIS and allowing a quantitative comparison. As an example, we report the results of a comparison run over three sites and 171 different dates, which can be found in Ref. [105]. These results show that conventional SU methods do not achieve a good snow cover fraction estimation with a typical error around 30% on the SCF maps. These errors give an evidence of the limitations of generic SU approaches which could be associated to the spectral variability of snow in an area. As a matter of fact, snow does not appear always with the same characteristics. For example, as in spring and summer, a snow cap can be covered by debris like dust, little rocks, or sand. In this case, snow spectra can vary within the scene (due to different kinds of impurities which are present) and with respect to reference spectra of “pure” snow as available from spectral libraries. For these reasons, it is important to consider more attentively the spectral variability of snow. Recent approaches in SU have been proposed to handle spectral variability of endmembers. An example is the so-called Extended Linear Mixing Model (ELMM [106]) in which a multiplicative scalar coefficient per endmember and per pixel can compensate to spectral variability. This has proven to be effective especially when spectral variability is due to scene topography. When

390 SECTION j III Application fields

considered in SCF estimation, as reported in the tests in Ref. [105], ELMM outperforms conventional SU approaches that do not take into account spectral variability. However, some limitations still exist, mainly due to the limited number of spectral bands available in the MODIS acquisition which can prevent the discrimination of different types of snow.

4.3 Open challenges in snow cover mapping The current main challenge in the application of hyperspectral images for snow cover mapping is the lack of periodic acquisitions due to the fact that no hyperspectral satellite sensor is available at the moment with spatial, spectral, and temporal resolutions adequate for the estimation of SCF maps. MODIS or other multispectral sensor can provide suitable spatial and temporal resolutions in the acquisitions, but the reduced number of spectral bands can be a limiting factor for coping with the spectral variability of snow (as reported in the previous section). The recent launch of the Sentinel-2 constellation featuring 10 bands in visible and near-infrared can produce higher spatial resolution SCF maps. However the revisit time of 5 days can be a limitation for the monitoring of rapidly evolving snow covers. It can be interesting to explore the possibilities provided by the new generations of geostationary sensors like GOES or HIMAWARI. They provide acquisitions with a spatial resolution of 500 m in the visible spectrum, with a very high temporal resolution in the visible, near-infrared, and infrared domains. Recent advances in data fusion can be beneficial for combining acquisitions on the same area from sensors with different spatial, spectral, and temporal resolutions. This would allow to retrieve snow cover from multiple acquisitions with potentially a more precise spatial, spectral, and temporal resolutions.

5. Vegetation 5.1 Context The ecological functions of terrestrial vegetation are fundamental for many processes and services occurring from local to global scale, including climate and water resources regulation, biogeochemical cycles, habitats for fauna, and food diversity, and security. These ecological functions are tightly linked to biodiversity. The decline of biodiversity has been well documented during the past decades, and attempts to reduce this decline did not succeed despite the awareness of the international community [107]. Multiple drivers can be blamed for this biodiversity decline: land use and cropping practice changes, increased pollution levels, changes in the biogeochemical cycles, climate alterations, and spread of plant diseases and exotic species [108,109]. These drivers show complex interactions, including feedback effects between biodiversity loss, climate change, and habitat degradation [110]. This results in a decreased ability of natural ecosystems and anthropogenic areas to maintain their ecological functions and to provide goods and services for society, with strong negative impact on human well-being [111].

Applications in remote sensingdnatural landscapes Chapter j 3.1

391

In response to this global issue, the Convention on Biological Diversity was adopted with the objectives to develop a strategic plan for the conservation, restoration, and sustainable use of biological diversity, based on various indicators. The implementation of this plan is under the responsibility of the Group on Earth Observation Biodiversity Observation Network (GEO BON) to ensure the development of efficient monitoring systems, combining appropriate monitoring tools focusing on relevant indicators of ecosystem processes [112]. Remote sensing is a crucial tool for the monitoring of taxonomic and functional diversity and so-called essential biodiversity variables [113e116]: it is repeatable, consistent, borderless, and scale independent. The link between in situ measurements of biodiversity and remotely sensed data remains challenging [117]: the concept of biodiversity is multidimensional; encompassing taxonomic, functional, phylogenetic, and genetic dimensions; and interactions among these dimensions. Leaf chemical and structural traits are pivotal for the study of these multiple dimensions of biodiversity: high diversity of plant communities leads to heterogeneity in chemical and structural traits [118], at local scale, and multiple biotic and abiotic controls also influence these traits at larger spatial scale and over time, filtering plant functions and strategies within and among species [119,120]. Therefore, monitoring leaf traits can provide information on complex processes driving the response of ecosystem functions to multiple environmental filters [121]. In return, leaf traits also interact with ecosystem functioning through processes such as primary production, litter decomposition, carbon and nutrient cycling. Spectroscopy proved to be a particularly efficient nondestructive technique for the estimation of leaf chemical and structural traits from their optical properties [122]. Various techniques based on statistical and machine learning methods [123e126], physical modeling [125,127e129], or the combination of both types of approaches [130] were used to estimate leaf pigment content (chlorophylls, carotenoids, anthocyanins), equivalent water thickness, leaf mass per area (LMA), and additional chemical constituents contents such as phosphorus and nitrogen. These traits directly influence absorption, reflection, and transmission of the electromagnetic energy when interacting with leaf. This “chemical fingerprint” directly influences the spectral signature of a leaf, and both type of information can be used for higher level applications, such as species discrimination [131e134], or discrimination of hierarchical organization within genus to capture phenotypic variations among taxa [135]. Therefore, chemical and spectral diversities are relevant proxies for taxonomic and phenotypic diversity. Leaf aging is also a strong driver of changes in leaf traits and optical properties [136], which means that spectral and chemical discrimination among species, phenotypes, or leaf development stage independently is possible, but that simultaneous discrimination among multiple drivers increases the complexity of relating optical properties to a combination of drivers and results in lower discrimination ability [137,138].

392 SECTION j III Application fields

At the canopy level, imaging spectroscopy also showed strong potential for the monitoring of various components of biodiversity. Direct upscaling of methods validated at the leaf scale to canopy scale remains challenging due to the large number of factors interacting with the electromagnetic energy, including leaf traits, but also canopy structure, geometry of acquisition, and environmental conditions. Recently, Feilhauer et al. [139] tested whether remotely sensed traits provide the required level of detail to trace the trait plasticity in response to long-term drought stress for wetland vegetation. They used physical model inversion to estimate LMA from imaging spectroscopy and concluded on the relevance of remotely sensed traits for ecological monitoring. Although promising, approaches based on physical modeling are usually limited by the capacity to properly describe the complexity of vegetation structure and leaf traits, and result in simplistic representation [132,140]. Therefore, statistical and machine learning algorithms are usually preferred when mapping canopy chemical traits in heterogeneous ecosystems [141e146], but these approaches require very intensive field data collection, and the generalization ability of the regression models calibrated for one sensor and one type of ecosystem still needs more investigation. Canopy chemical traits derived from imaging spectroscopy hold strong potential for characterizing biodiversity, monitoring ecosystem functions, and understanding drivers of change of these functions. However, existing approaches show limitations due to the important amount of data required to train regression algorithms or run physical models with appropriate level of details. The moderate number of traits which can be estimated, and the uncertainty associated to these estimations also contributes to these limitations. Efforts are made to overcome these limitations [141], and alternative approaches have been developed in order to directly relate spectral information to biodiversity. The main advantage is that the full spectral information can be used, and the main drawback is that it increases the difficulty to perform physical and ecological interpretation of the relationships between spectroscopy and biodiversity. Mapping biodiversity directly from optical data requires defining a conceptual framework to understand how diversity can be expressed in a remote sensing point of view. Ustin and Gamon [147] first introduced plant optical types as a concept defining the capacity of optical remote sensing information to distinguish functional type among plants, and Feilhauer et al. [148] broadened the concept to “optical traits,” corresponding to optically effective plant properties measured with spectroscopy. Similar concepts have been explored in combination with the spectral variation hypothesis (SVH) [149,150], in order to map various components of biodiversity from imaging spectroscopy acquired over various ecosystems, including tropical forests [151], savannah [152,153], and grasslands [154e156]. The method developed by Fe´ret and Asner [151] is proposed as illustration for biodiversity mapping from airborne imaging spectroscopy in tropical forests.

Applications in remote sensingdnatural landscapes Chapter j 3.1

393

5.2 Example of application: mapping biodiversity in tropical ecosystems The method developed by Fe´ret and Asner [151] proposes a single framework to compute both a and b components of biodiversity [157,158]. It is built upon two concepts previously introduced, the SVH proposed by Palmer et al [149], and the plant optical types proposed by Ustin and Gamon [147]. The SVH states that the spatial variability measured from remotely sensed information, usually corresponding to the spectral reflectance for optical sensors, is expected to be related to environmental heterogeneity and could therefore be used as a powerful proxy of species diversity. SVH has been tested in different situations [150], and conclusions show that the performances of this approach depend on several factors, including the instrumental characteristics (spectral, spatial, and temporal resolutions), the type of vegetation investigated, and the metrics derived from remotely sensed information to estimate spectral heterogeneity. Plant optical types refer to the capacity of sensors to measure signal aggregating information about vegetation structure, phenology, biochemistry, and physiology. Therefore, this concept is also tightly linked to the performances of the sensor and finds particular echo with the increasing use of high spatial resolution imaging spectroscopy for the estimation and identification of multiple vegetation properties: pixels corresponding to the same species in an image are expected to share similar plant optical types, or at least stronger similarity than pixels corresponding to different species. In this case, the spatial resolution is assumed to be high enough to distinguish among species. In the method developed by Fe´ret and Asner [151], a-diversity corresponds to an indicator of species diversity for a given spatial unit and can be expressed as species richness, Shannon index, or Simpson index; the diversity metric defined as b-diversity in this framework corresponds to the dissimilarity among pairs of spatial units and is expressed as a dissimilarity metric, such as the BrayeCurtis dissimilarity. The method is based on the following rationales: (1) Exhaustive tree crown delineation and species identification is unrealistic in megadiverse ecosystems, even from imaging spectroscopy; (2) A majority of the pixels from high spatial resolution imaging spectroscopy (metric resolution) can be assigned to individual species, as most tree crowns emerging from the canopy are several times larger than pixels; (3) Unsupervised clustering of properly preprocessed imaging spectroscopy data may result in clustering species or groups of species; (4) Following the idea that each pixel corresponds to one individual, and one cluster corresponds to one species, diversity indices can be computed from imaging spectroscopy, the same way they are computed from inventories. This rationale is based on strong hypotheses and simplifications: variability in individual tree size is not accounted for in the image, which means that pixels are expected to be “pure species,” and pixels from the same cluster spatially connected are seen as a group of individuals instead of a unique individual; finally, the number of clusters defined for the k-means clustering is not set based on prior information on species diversity.

394 SECTION j III Application fields

Data preprocessing is crucial and aims at maximizing discrimination among species in preparation for unsupervised clustering. Several preprocessing stages are then required. First, all nonvegetated pixels are excluded. Nonvegetated pixels would strongly increase heterogeneity, and assigning clusters to such pixels would decrease the capacity to distinguish among species or groups of species. Then, a normalization based on continuum removal is applied to each pixel and over the full spectral domain, in order to minimize spectral variations induced by illumination, which is the main source of variation for high spatial resolution imaging spectroscopy acquired over forested areas. Finally, a PCA is performed on the continuum removed spectral data. The components including individual-specific information are the components of interest. They can be identified after visual inspection or automated routines, if initial data show sufficient SNR and spatial resolution. Once a limited number of components

FIGURE 10 Example of biodiversity mapping results throughout the CICRA site located in lowland Amazonia. Panels are as follows: (A) natural color composite image from the CAO visible-to-shortwave infrared (VSWIR) imaging spectrometer; (B) a-diversity based on the Shannon index; and (C) b-diversity based on BrayeCurtis dissimilarity (no color scale is applicable, and a larger BrayeCurtis dissimilarity between two plots corresponds to larger differences in color in the RGB space between the two corresponding pixels). Results from J.B. Fe´ret, G.P. Asner, Mapping tropical forest canopy diversity using high-fidelity imaging spectroscopy. Ecological Applications 24 (6) (2014) 1289e1296.

Applications in remote sensingdnatural landscapes Chapter j 3.1

395

have been selected, k-means clustering is then applied to a subset of pixel, and each pixel in the image is labeled based on the closest centroid. This method was tested successfully on a limited number of situations and types of ecosystems. Fig. 10 illustrates the output of the method for a- and bdiversity mapping over a site in Amazonian rainforest, based on CAO-AToMS imaging spectroscopy [159]. A generic parameterization of the method showed robust performances for a-diversity mapping across space and time, but mapping b-diversity across large spatial scales using images acquired during different airborne campaign remains challenging, which leads to unsolved problem when considering operational regional mapping. In the perspective of global monitoring of biodiversity, and based on the unprecedented remote sensing capacity allowed by the Copernicus program, including the Sentinel-2 multispectral satellites, several other challenges are foreseen and currently investigated. The influence of decreased spatial and spectral resolution on the ability to properly differentiate ecologically meaningful spectral species across landscapes and over regions will need to be investigated. The application of this concept beyond tropical forests and savannah ecosystems should also be investigated, as it may not hold when applied on moderately diverse ecosystems or systems with individuals with lower than metric dimensions.

5.3 Open challenges in monitoring vegetation Imaging spectroscopy proved its strong potential for the monitoring of vegetation properties, including leaf traits, plant and ecosystem functions, and taxonomic diversity. However, the predictive power of this technology is usually superior to the explanatory power. Better performances reported when using spectral information instead of plant traits information for the estimation of some components of biodiversity at both leaf scale [135] and canopy scale [160] are a good illustration. In this situation, and given the strong correlations occurring among multiple plant traits [161,162], differentiating causation (variations in leaf traits influencing optical properties) from correlation (variation in optically inactive leaf constituent predicted from spectroscopic measurements because of its correlation with an optically active constituent) is not straightforward. The predictive power of statistical and machine learning models may be overestimated for many other reasons, including bias or lack of representativity in training data and differences in performances among sensors [128]. Therefore, improved robustness of predictive models requires open access data sets including extensive field data and imaging spectroscopy data, as well as intercomparison exercises of these predictive models applied on independent data sets. Leaf traits available from physical models are currently limited to constituents showing the strongest absorption features in the optical domain. Improved physical understanding of the relationship between leaf traits and leaf optical properties is also needed in order to better understand the potential and limitations of spectroscopy for the estimation of leaf traits at both leaf and canopy

396 SECTION j III Application fields

FIGURE 11 Simulation of airborne optical imaging in a tropical forest from French Guiana (Paracou). The simulation is based on the integration of airborne LiDAR and field spectroscopy into the 3D radiative transfer model DART [163]. Three-dimensional mockups were computed from airborne LiDAR cloud points using AMAPvox [164], and leaf optical properties corresponding to sampled trees (delineated in red) were assigned to leaf elements from all voxels on the vertical column of the mockup. A generic set of leaf optical properties was applied to all trees with undocumented leaf optical properties using the PROSPECT leaf model [127]. Left: original image (red ¼ 640 nm; green ¼ 549 nm; blue ¼ 458 nm); center: simulation using a turbid representation for leaf elements; right: simulation using triangle approach for leaf elements.

scale, and test the explanatory power for constituents such as micronutrients. Improved physical understanding of the interactions between leaf traits and electromagnetic energy combined with enhanced capacity to accurately describe the 3D structure of complex ecosystems based on LiDAR information now allows taking advantage of 3D radiative transfer models such as DART [163]. Fig. 11 illustrates the comparison between airborne imaging spectroscopy and its simulated counterpart with the DART model, based on LiDAR acquisitions and leaf spectroscopic measurements for a series of delineated trees. Forthcoming hyperspectral satellite missions including PRISMA [165], HISUI [166], and EnMAP [167] will soon open new perspectives for applications in environmental monitoring based on imaging spectroscopy, including ecosystems functions and biodiversity mapping.

6. Discussion and conclusion This chapter presented four thematic applications related to the use of hyperspectral imagery for the monitoring of natural landscapes. By looking at the case studies reported as examples for each application, it is possible to see how hyperspectral imagery is a privileged tool for characterize natural environments. From the analysis of hyperspectral acquisitions, it is possible to get access to information on the surveyed scene such as the detection of some target materials in the area of interest, the estimation of subpixel proportions of materials, the thematic classification of land covers in a supervised or unsupervised way, and the estimation of some physical characteristics of the scene. In terms of methodology used for the analysis of hyperspectral acquisitions, it is possible to notice some similarities in terms of methodologies that are

Applications in remote sensingdnatural landscapes Chapter j 3.1

397

employed. For example, SU stands out as one of the most used analytical approaches allowing one to access subpixel characteristics in images such as material abundances, which are useful in the contexts of planetary surface investigation, estimation of snow cover maps, and mapping coastal areas. Scene classification is another common approach in which, the high spectral resolution of hyperspectral acquisitions provides land cover maps giving the spatial arrangement of materials in the scene, as in the case study on coastal areas. Moreover, after identification, one can also extract some further information on the morphology of some land cover classes via image processing (e.g., scale and shape of objects belonging to a given class) as it was reported in the analysis of dark spots on Mars images. The combination of hyperspectral acquisitions with those of different sensors through image fusion is another recurrent method for coping with limitations of the hyperspectral acquisitions in terms of spatial or temporal resolutions. The analysis of hyperspectral time series is another interesting area of research which is currently not much explored mainly due to the current lack of hyperspectral satellites which could provide hyperspectral time series for Earth observation. However, hyperspectral time series on Mars has proven the potential of having periodic acquisitions, for example, for change detection. In all application scenarios, taking into account a physical model is of utmost importance for providing a priori information that can favor the selection of physical meaningful solutions by the analytical tools employed. Moreover, physical models are also needed for the retrieval of the pixel reflectance (e.g., by atmospheric correction or by compensating the presence of a water column for accessing reflectance of the seabed), which is a common starting point for most of the subsequent analyses. In terms of challenges, the ones that seems more recurrent across the four thematic applications are (1) the relatively low spatial resolution of current hyperspectral acquisitions that can be a limiting factor for characterizing objects of interest in the scene; (2) the lack of hyperspectral satellites able to ensure periodic acquisitions; (3) the development of analysis which can better include (more accurate) physical models and ancillary information from a scene.

Acknowledgments The collaborative writing of this chapter has been made possible, thanks to animation and outreach activities organized by the IEEE Geoscience and Remote Sensing (GRSS) French Chapter, the Hyperspectral Group of the French Society of Photogrammetry and Remote Sensing (Groupe Hyperspectrale de la Societe´ Franc¸aise de Photogrammetrie et Te´le´de´tection) and the Action Imagerie Hyperspectrale (ImHyp) of the GdR MaDICS, which fostered exchanges and collaborations between coauthors that eventually led to this work. The research presented in this chapter was partially supported by grants from Labex OSUG@2020 (Investissements d’avenir ANR10 LABX56), the Programme National de Te´le´de´tection Spatiale (PNTS), grant n. PNTS-2016-03, the ARC3 program of the Region Rhone Alpes and support from the “Institut National des Sciences de l’Univers” (INSU), the “Center National de la Recherche Scientifique” (CNRS), “Center National d’Etude Spatiale”

398 SECTION j III Application fields (CNES), the “Programme National de Plane´tologie,” MEX/OMEGA Program, the “Agence Nationale pour la Recherche”, through the funding of the ANR JCJC project ‘BioCop’ (ANR-17-CE32-0001-01), and the TOSCA funding grant program (project ‘HyperTropik’).

References [1] D.G. Manolakis, R.B. Lockwood, T.W. Cooley, Hyperspectral Imaging Remote Sensing: Physics, Sensors, and Algorithms, Cambridge University Press, 2016. [2] W. Emery, A. Camps, Introduction to Satellite Remote Sensing: Atmosphere, Ocean, Land and Cryosphere Applications, Elsevier, 2017. [3] J.-P. Bibring, M. Combes, Y. Langevin, A. Soufflot, C. Cara, P. Drossart, T. Encrenaz, S. Erard, O. Forni, B. Gondet, L. Ksanfomalfty, E. Lellouch, P. Masson, V. Moroz, F. Rocard, J. Rosenqvist, C. Sotin, Results from the ism experiment, Nature 341 (6243) (1989) 591e593. [4] P.R. Christensen, D.L. Anderson, S.C. Chase, R.N. Clark, H.H. Kieffer, M.C. Malin, J.C. Pearl, J. Carpenter, N. Bandiera, F. Gerald Brown, S. Silverman, Thermal emission spectrometer experiment: Mars observer mission, Journal of Geophysical Research: Planets 97 (E5) (1992) 7719e7734. [5] R.W. Carlson, P.R. Weissman, W.D. Smythe, J.C. Mahoney, Near-infrared mapping spectrometer experiment on galileo, Space Science Reviews 60 (1) (1992) 457e502. [6] G. Vane, R.O. Green, T.G. Chrien, H.T. Enmark, E.G. Hansen, W.M. Porter, The airborne visible/ infrared imaging spectrometer (AVIRIS), Remote Sensing of Environment 44 (1993) 127e143. [7] A. Coradini, F. Capaccioni, P. Drossart, A. Semery, G. Arnold, U. Schade, F. Angrilli, M. A Barucci, G. Bellucci, G. Bianchini, J. P Bibring, A. Blanco, M. Blecka, D. BockeleeMorvan, R. Bonsignori, M. Bouye, E. Bussoletti, M. T Capria, R. Carlson, U. Carsenty, P. Cerroni, L. Colangeli, M. Combes, M. Combi, J. Crovisier, M. Dami, M. C DeSanctis, A. M DiLellis, E. Dotto, T. Encrenaz, E. Epifani, S. Erard, S. Espinasse, A. Fave, C. Federico, U. Fink, S. Fonti, V. Formisano, Y. Hello, H. Hirsch, G. Huntzinger, R. Knoll, D. Kouach, W. H Ip, P. Irwin, J. Kachlicki, Y. Langevin, G. Magni, T. McCord, V. Mennella, H. Michaelis, G. Mondello, S. Mottola, G. Neukum, V. Orofino, R. Orosei, P. Palumbo, G. Peter, B. Pforte, G. Piccioni, J. M Reess, E. Ress, B. Saggin, B. Schmitt, Stefanovitch, A. Stern, F. Taylor, D. Tiphene, G. Tozzi, Virtis : an imaging spectrometer for the rosetta mission, Planetary and Space Science 46 (9e10) (1998) 1291e1304. [8] J.S. Pearlman, P.S. Barry, C.C. Segal, J. Shepanski, D. Beiso, S.L. Carman, Hyperion, a space-based imaging spectrometer, Geoscience and Remote Sensing, IEEE Transactions on 41 (6) (2003) 1160e1173. [9] J.-P. Bibring, A. Soufflot, M. Berthe´, Y. Langevin, B. Gondet, P. Drossart, M. Bouye´, M. Combes, P. Puget, A. Semery, G. Bellucci, V. Formisano, V. Moroz, V. Kottsov, G. Bonello, S. Erard, O. Forni, A. Gendrin, N. Manaud, F. Poulet, G. Poulleau, T. Encrenaz, T. Fouchet, R. Melchiori, F. Altieri, N. Ignatiev, D. Titov, L. Zasova, A. Coradini, F. Capacionni, P. Cerroni, S. Fonti, N. Mangold, P. Pinet, B. Schmitt, C. Sotin, E. Hauber, H. Hoffmann, R. Jaumann, U. Keller, R. Arvidson, J. Mustard, F. Forget, OMEGA: Observatoire pour la Mine´ralogie, l’Eau, les Glaces et l’Activite´, ESA SP-1240: Mars Express: the Scientific Payload, 2004, pp. 37e49. [10] R. Brown, K. Baines, G. Bellucci, J.-P. Bibring, B. Buratti, F. Capaccioni, P. Cerroni, R. Clark, A. Coradini, D. Cruikshank, P. Drossart, V. Formisano, R. Jaumann, Y. Langevin, D. Matson, T. Mccord, V. Mennella, E. Miller, R. Nelson, P. Nicholson, B. Sicardy, C. Sotin, The cassini visual and infrared mapping spectrometer (VIMS) investigation, in: The Cassini-Huygens Mission, 2004, pp. 111e168.

Applications in remote sensingdnatural landscapes Chapter j 3.1

399

[11] S. Murchie, R. Arvidson, P. Bedini, K. Beisser, J.-P. Bibring, J. Bishop, J. Boldt, P. Cavender, T. Choo, R.T. Clancy, E.H. Darlington, D. Des Marais, R. Espiritu, D. Fort, R. Green, E. Guinness, J. Hayes, C. Hash, K. Heffernan, J. Hemmler, G. Heyler, D. Humm, J. Hutcheson, N. Izenberg, R. Lee, J. Lees, D. Lohr, E. Malaret, T. Martin, J.A. McGovern, P. McGuire, R. Morris, J. Mustard, S. Pelkey, E. Rhodes, M. Robinson, T. Roush, E. Schaefer, G. Seagrave, F. Seelos, P. Silverglate, S. Slavney, M. Smith, W.-J. Shyong, K. Strohbehn, H. Taylor, P. Thompson, B. Tossman, M. Wirzburger, M. Wolff, Compact reconnaissance imaging spectrometer for Mars (CRISM) on Mars reconnaissance orbiter (MRO), Journal of Geophysical Research 112 (E5) (2007). E05S03. [12] G. Piccioni, P. Drossart, E. Suetta, M. Cosi, E. Amannito, A. Barbis, R. Berlin, A. Bocaccini, G. Bonello, M. Bouye´, F. Capaccioni, G. Cherubini, M. Dami, O. Dupuis, A. Fave, G. Filacchione, Y. Hello, F. Henry, S. Hofer, G. Huntzinger, R. Melchiorri, J. Parisot, C. Pasqui, G. Peter, C. Pompei, J.M. Re`ess, A. Semery, A. Soufflot, A. Adriani, F. Angrilli, G. Arnold, K. Baines, G. Bellucci, J. Benkhoff, B. Bezard, J.-P. Bibring, A. Blanco, M.I. Blecka, R. Carlson, A. Coradini, A. Di Lellis, T. Encrenaz, S. Erard, S. Fonti, V. Formisano, T. Fouchet, R. Garcia, R. Haus, J. Helbert, N.I. Ignatiev, P. Irwin, Y. Langevin, S. Lebonnois, M.A. Lopez Valverde, D. Luz, M. Marinangeli, V. Orofino, A.V. Rodin, M.C. Roos-Serote, B. Saggin, A. Sanchez-Lavega, B.M. Stam, F. Taylor, D. Titov, G. Visconti, M. Zambelli, VIRTIS: The Visible and Infrared Thermal Imaging Spectrometer, ESA Special Publication, 1295, 2007. [13] C.M. Pieters, B. Joseph, B. Buratti, A. Chatterjee, R. Clark, T. Glavich, R. Green, J. Head III, Peter Isaacson, E. Malaret, et al., The moon mineralogy mapper (M3) on Chandrayaan-1, Current Science 96 (4) (2009) 500e505. [14] M.C. Sanctis, A. Coradini, E. Ammannito, G. Filacchione, M.T. Capria, S. Fonte, G. Magni, A. Barbis, A. Bini, M. Dami, I. Ficai-Veltroni, G. Preti, The VIR spectrometer, Space Science Reviews 163 (1e4) (2011) 329e369. [15] D.C. Reuter, S. Alan Stern, J. Scherrer, D.E. Jennings, J.W. Baer, J. Hanley, L. Hardaway, L. Allen, S. McMuldroch, J. Moore, C. Olkin, R. Parizek, H. Reitsma, D. Sabatke, J. Spencer, J. Stone, T. Henry, J. Van Cleve, G.E. Weigle, L.A.Y. Ralph, A visible/infrared imager for the new horizons pluto/kuiper belt mission, Space Science Reviews 140 (1e4) (2008) 129e154. [16] E. Flamini, F. Capaccioni, L. Colangeli, G. Cremonese, A. Doressoundiram, J.L. Josset, Y. Langevin, S. Debei, M.T. Capria, M.C. De Sanctis, L. Marinangeli, M. Massironi, E. Mazzotta Epifani, G. Naletto, P. Palumbo, P. Eng, J.F. Roig, A. Caporali, V. Da Deppo, S. Erard, C. Federico, O. Forni, M. Sgavetti, G. Filacchione, L. Giacomini, G. Marra, E. Martellato, M. Zusi, M. Cosi, C. Bettanini, L. Calamai, M. Zaccariotto, L. Tommasi, M. Dami, J. Ficai Veltroni, F. Poulet, Y. Hello, SIMBIO-SYS: the spectrometer and imagers integrated observatory system for the BepiColombo planetary orbiters, Planetary and Space Science 58 (1e2) (2010) 125e143. Comprehensive Science Investigations of Mercury: The scientific goals of the joint ESA/JAXA mission BepiColombo. [17] H. Hiesinger, J. Helbert, The mercury radiometer and thermal infrared spectrometer (MERTIS) for the BepiColombo mission, Planetary and Space Science 58 (1e2) (2010) 144e165. Comprehensive Science Investigations of Mercury: The scientific goals of the joint ESA/JAXA mission BepiColombo. [18] Y. Langevin, G. Piccioni, MAJIS (Moons and Jupiter Imaging Spectrometer) for JUICE: Objectives for the Galilean Satellites, vol. 8, 2013, pp. EPSC2013eEPSC2548. [19] P. Drossart, G. Piccioni, J.C. Ge´rard, M.A. Lopez-Valverde, A. Sanchez-Lavega, L. Zasova, R. Hueso, F.W. Taylor, B. Be´zard, A. Adriani, F. Angrilli, G. Arnold,

400 SECTION j III Application fields

[20]

[21]

[22]

[23]

[24]

[25] [26]

[27]

[28]

[29] [30]

K.H. Baines, G. Bellucci, J. Benkhoff, J.P. Bibring, A. Blanco, M.I. Blecka, R.W. Carlson, A. Coradini, A. Di Lellis, T. Encrenaz, S. Erard, S. Fonti, V. Formisano, T. Fouchet, R. Garcia, R. Haus, J. Helbert, N.I. Ignatiev, P. Irwin, Y. Langevin, S. Lebonnois, D. Luz, L. Marinangeli, V. Orofino, A.V. Rodin, M.C. Roos-Serote, B. Saggin, D.M. Stam, D. Titov, G. Visconti, M. Zambelli, C. Tsang, A dynamic upper atmosphere of venus as revealed by VIRTIS on venus express, Nature 450 (7170) (2007) 641e645. G. Piccioni, P. Drossart, A. Sanchez-Lavega, R. Hueso, F.W. Taylor, C.F. Wilson, D. Grassi, L. Zasova, M. Moriconi, A. Adriani, S. Lebonnois, A. Coradini, B. Be´zard, F. Angrilli, G. Arnold, K.H. Baines, G. Bellucci, J. Benkhoff, J.P. Bibring, A. Blanco, M.I. Blecka, R.W. Carlson, A. Di Lellis, T. Encrenaz, S. Erard, S. Fonti, V. Formisano, T. Fouchet, R. Garcia, R. Haus, J. Helbert, N.I. Ignatiev, P.G.J. Irwin, Y. Langevin, M.A. Lopez-Valverde, D. Luz, L. Marinangeli, V. Orofino, A.V. Rodin, M.C. Roos-Serote, B. Saggin, D.M. Stam, D. Titov, G. Visconti, M. Zambelli, South-polar features on venus similar to those near the North pole, Nature 450 (7170) (2007) 637e640. J.-P. Bibring, Y. Langevin, A. Gendrin, B. Gondet, F. Poulet, M. Berthe´, A. Soufflot, R. Arvidson, N. Mangold, J. Mustard, P. Drossart, Mars surface diversity as revealed by the OMEGA/Mars express observations, Science 307 (2005) 1576e1581. J.F. Mustard, S.L. Murchie, S.M. Pelkey, B.L. Ehlmann, R.E. Milliken, J.A. Grant, J.P. Bibring, F. Poulet, J. Bishop, E.N. Dobrea, L. Roach, F. Seelos, R.E. Arvidson, S. Wiseman, R. Green, C. Hash, D. Humm, E. Malaret, J.A. McGovern, K. Seelos, T. Clancy, R. Clark, D.D. Marais, N. Izenberg, A. Knudson, Y. Langevin, T. Martin, P. McGuire, R. Morris, M. Robinson, T. Roush, M. Smith, G. Swayze, H. Taylor, T. Titus, M. Wolff, Hydrated silicate minerals on Mars observed by the Mars reconnaissance orbiter CRISM instrument, Nature 454 (7202) (2008) 305e309. Y. Langevin, F. Poulet, J.-P. Bibring, B. Schmitt, S. Doute´, B. Gondet, Summer evolution of the North polar cap of Mars as observed by OMEGA/Mars express, Science 307 (2005) 1581e1584. F. Montmessin, B. Gondet, J.-P. Bibring, Y. Langevin, P. Drossart, F. Forget, T. Fouchet, Hyperspectral imaging of convective CO2 ice clouds in the equatorial mesosphere of Mars, Journal of Geophysical Research (Planets) 112 (2007) 11. A.G. Davies, R.M.C. Lopes-Gautier, W.D. Smythe, R.W. Carlson, Silicate cooling model fits to Galileo NIMS data of volcanism on Io, Icarus 148 (2000) 211e225. S. Doute´, B. Schmitt, R. Lopes-Gautier, R. Carlson, L. Soderblom, J. Shirley, The Galileo NIMS Team, Mapping SO2 frost on Io by the modeling of NIMS hyperspectral images, Icarus 149 (2001) 107e132. K.P. Hand, R.W. Carlson, Europa’s surface color suggests an ocean rich with sodium chloride, Geophysical Research Letters (2015) 3174e3178. https://doi.org/10.1002/ 2015GL063559. S. Rodriguez, S. Le Moue´lic, P. Rannou, G. Tobie, K.H. Baines, J.W. Barnes, C.A. Griffith, M. Hirtzig, K.M. Pitman, C. Sotin, R.H. Brown, B.J. Buratti, R.N. Clark, P.D. Nicholson, Global circulation as the main source of cloud activity on Titan, Nature 459 (2009) 678e682. ´ da´mkovics, Titan’s surface and atmosphere, Icarus 270 A.G. Hayes, J.M. Soderblom, M. A (2016). B. Schmitt, S. Philippe, W.M. Grundy, D.C. Reuter, R. Coˆte, E. Quirico, S. Protopapa, L.A. Young, R.P. Binzel, J.C. Cook, D.P. Cruikshank, C.M. Dalle Ore, A.M. Earle, K. Ennico, C.J.A. Howett, D.E. Jennings, I.R. Linscott, A.W. Lunsford, C.B. Olkin, A.H. Parker, J.W. Parker, K.N. Singer, J.R. Spencer, J.A. Stansberry, S.A. Stern,

Applications in remote sensingdnatural landscapes Chapter j 3.1

[31]

[32]

[33]

[34] [35]

[36]

[37]

[38]

[39]

[40]

[41]

[42]

[43]

[44]

401

C.C.C. Tsang, A.J. Verbiscer, H.A. Weaver, Physical state and distribution of materials at the surface of Pluto from New Horizons LEISA imaging spectrometer, Icarus 287 (2017) 229e260. F. Schmidt, S. Doute´, B. Schmitt, WAVANGLET: an efficient supervised classifier for hyperspectral images, Geoscience and Remote Sensing, IEEE Transactions on 45 (5) (2007) 1374e1385. J.-P. Combe, S. Le Moue´lic, C. Sotin, A. Gendrin, J.F. Mustard, L. Le Deit, P. Launeau, J.P. Bibring, B. Gondet, Y. Langevin, P. Pinet, The OMEGA Science Team, Analysis of OMEGA/Mars express data hyperspectral data using a multiple-endmember linear spectral unmixing model (MELSUM): methodology and first results, Planetary and Space Science 56 (2008) 951e975. K.E. Themelis, F. Schmidt, O. Sykioti, A.A. Rontogiannis, K.D. Koutroumbas, I.A. Daglis, On the unmixing of mex/omega hyperspectral data, Planetary and Space Science 68 (1) (2012) 34e41. F. Schmidt, M. Legendre, S. Le Moue¨lic, Minerals detection for hyperspectral images using adapted linear unmixing: LinMin, Icarus 237 (2014) 61e74. S. Doute´, B. Schmitt, Y. Langevin, J.-P. Bibring, F. Altieri, G. Bellucci, B. Gondet, F. Poulet, The MEX OMEGA Team, South Pole of Mars: nature and composition of the icy terrains from Mars Express OMEGA observations, Planetary and Space Science 55 (2007) 113e133. S. Erard, P. Drossart, G. Piccioni, Multivariate analysis of visible and infrared thermal imaging spectrometer (VIRTIS) venus express nightside and limb observations, Journal of Geophysical Research 114 (2009). S. Moussaoui, H. Hauksdo´ttir, F. Schmidt, C. Jutten, J. Chanussot, D. Brie, S. Doute´, J.A. Benediktsson, On the decomposition of Mars hyperspectral data by ICA and Bayesian positive source separation, Neurocomputing 71 (10e12) (2008) 2194e2208. M.J. Wolff, M.D. Smith, R.T. Clancy, R. Arvidson, M. Kahre, F. Seelos, S. Murchie, H. Savija¨ rvi, Wavelength dependence of dust aerosol single scattering albedo as observed by the compact reconnaissance imaging spectrometer, Journal of Geophysical Research 114 (E2) (2009). E00D04. X. Ceamanos, S. Doute´, J. Fernando, F. Schmidt, P. Pinet, A. Lyapustin, Surface reflectance of Mars observed by CRISM/MRO: 1. Multi-angle approach for retrieval of surface reflectance from CRISM observations (MARS-ReCO), Journal of Geophysical Research (Planets) 118 (2013) 514e533. J. Fernando, F. Schmidt, X. Ceamanos, P. Pinet, S. Doute´, Y. Daydou, Surface reflectance of Mars observed by CRISM/MRO: 2. Estimation of surface photometric properties in Gusev crater and Meridiani Planum, Journal of Geophysical Research (Planets) 118 (2013) 534e559. J. Fernando, F. Schmidt, S. Doute´, Martian surface microtexture from orbital CRISM multi-angular observations: a new perspective for the characterization of the geological processes, Planetary and Space Science 128 (2016) 30e51. E. Gardin, P. Allemand, C. Quantin, P. Thollot, Defrosting, dark flow features, and dune activity on Mars: example in russell crater, Journal of Geophysical Research 115 (E6) (2010). E06016. X. Huang, L. Zhang, Morphological building/shadow index for building extraction from high-resolution imagery over urban areas, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 5 (1) (2012) 161e172. X. Ceamanos, S. Doute, B. Luo, F. Schmidt, G. Jouannic, J. Chanussot, Intercomparison and validation of techniques for spectral unmixing of hyperspectral images: a planetary

402 SECTION j III Application fields

[45] [46] [47] [48] [49] [50] [51]

[52]

[53]

[54]

[55]

[56]

[57]

[58]

[59]

[60] [61]

case study, Geoscience and Remote Sensing, IEEE Transactions on 49 (11) (2011) 4341e4358. https://doi.org/10.1109/TGRS.2011.2140377. F. Andrieu, S. Doute, F. Schmidt, B. Schmitt, Radiative transfer model for contaminated rough slabs, Applied Optics 54 (31) (2015) 9228. F. Andrieu, F. Schmidt, B. Schmitt, S. Doute´, O. Brissaud, Retrieving the characteristics of slab ice covering snow by remote sensing, The Cryosphere 10 (5) (2016) 2113e2128. F. Andrieu, F. Schmidt, S. Doute´, E. Chassefie`re, Ice state evolution during spring in richardson crater, Mars, Icarus 315 (2018) 158e173. UNEP-WCMC: Millennium Ecosystem Assessment, Coastal Vector Digital Data, Island Press, 2005. UNEP-WCMC: Millennium Ecosystem Assessment, Inland Water Raster Digital Data, Island Press, 2005. C.M. Duarte, W.C. Dennison, R.J.W. Orth, T.J.B. Carruthers, The charisma of coastal ecosystems: addressing the imbalance, Estuaries and Coasts 31 (2) (2008) 233e238. S. Jay, M. Guillaume, A. Minghelli, Y. Deville, C. Malik, B. Lafrance, V. Serfaty, Hyperspectral remote sensing of shallow waters: considering environmental noise and bottom intra-class variability for modeling and inversion of water reflectance, Remote Sensing of Environment 200 (2017) 352e367. E.J. Botha, V. E Brando, J. M Anstee, A.G. Dekker, Stephen Sagar, Increased spectral resolution enhances coral detection under varying water conditions, Remote Sensing of Environment 131 (2013) 247e261. J. Hedley, C. Roelfsema, B. Koetz, P. Stuart, Capability of the sentinel 2 mission for tropical coral reef mapping and coral bleaching detection, Remote Sensing of Environment 120 (2012) 145e155. E.J. Hochberg, M.J. Atkinson, S. Andre´foue¨t, Spectral reflectance of coral reef bottomtypes worldwide and implications for coral reef remote sensing, Remote Sensing of Environment 85 (2) (2003) 159e173. T. Kutser, A.G. Dekker, W. Skirving, Modeling spectral discrimination of great barrier reef benthic communities by remote sensing instruments, Limnology and Oceanography 48 (1part2) (2003) 497e510. M. Reichstetter, P. Fearns, S. Weeks, L. McKinna, C. Roelfsema, M. Furnas, Bottom reflectance in ocean color satellite remote sensing for coral reef environments, Remote Sensing 7 (12) (2015) 16756e16777. F. Kazemipour, P. Launeau, V. Me´le´der, Microphytobenthos biomass mapping using the optical model of diatom biofilms: application to hyperspectral images of Bourgneuf bay, Remote Sensing of Environment 127 (2012) 1e13. F. Kazemipour, V. Me´le´der, P. Launeau, Optical properties of microphytobenthic biofilms (MPBOM): biomass retrieval implication, Journal of Quantitative Spectroscopy and Radiative Transfer 112 (1) (2011) 131e142. A. Chennu, F. Paul, N. Volkenborn, M.A.A. Al-Najjar, F. Janssen, D. de Beer, L. Polerecky, Hyperspectral imaging of the microscale distribution and dynamics of microphytobenthos in intertidal sediments, Limnology and Oceanography: Methods 11 (10) (2013) 511e528. K.S. Schmidt, A.K. Skidmore, Spectral discrimination of vegetation types in a coastal wetland, Remote Sensing of Environment 85 (1) (2003) 92e108. S.K. Fyfe, Spatial and temporal variation in spectral reflectance: are seagrass species spectrally distinct? Limnology and Oceanography 48 (1part2) (2003) 464e479.

Applications in remote sensingdnatural landscapes Chapter j 3.1

403

[62] V.J. Hill, R.C. Zimmerman, W. Paul Bissett, H. Dierssen, D.D.R. Kohler, Evaluating light availability, seagrass biomass, and productivity using hyperspectral airborne remote sensing in Saint Joseph’s Bay, Florida, Estuaries and Coasts 37 (6) (2014) 1467e1489. [63] E.J. Hochberg, M.J. Atkinson, Spectral discrimination of coral reef benthic communities, Coral Reefs 19 (2) (2000) 164e171. [64] D.R. Mishra, S. Narumalani, D. Rundquist, M. Lawson, R. Perk, Enhancing the detection and classification of coral reef and associated benthic habitats: a hyperspectral remote sensing approach, Journal of Geophysical Research: Oceans 112 (C8) (2007). [65] G. Johnsen, Z. Volent, H. Dierssen, R. Pettersen, M. Van Ardelan, Fredrik Søreide, P. Fearns, L. Martin, M. Moline, Underwater hyperspectral imagery to create biogeochemical maps of seafloor properties, in: Subsea Optics and Imaging, Elsevier, 2013, pp. 508e535, 536e-540e. [66] C.D. Mobley, Light and Water: Radiative Transfer in Natural Waters, Academic press, 1994. [67] A.G. Dekker, S.R. Phinn, J. Anstee, B. Paul, V.E. Brando, B. Casey, P. Fearns, J. Hedley, W. Klonowski, Z.P. Lee, et al., Intercomparison of shallow water bathymetry, hydro-optics, and benthos mapping techniques in australian and caribbean coastal environments, Limnology and Oceanography: Methods 9 (9) (2011) 396e425. [68] Z. Lee, K.L. Carder, C.D. Mobley, R.G. Steward, J.S. Patch, Hyperspectral remote sensing for shallow waters. I. A semianalytical model, Applied Optics 37 (27) (1998) 6329e6338. [69] C.D. Mobley, L.K. Sundman, C.O. Davis, J.H. Bowles, T. Valerie Downes, R.A. Leathers, M.J. Montes, W. Paul Bissett, D.D.R. Kohler, R.P. Reid, et al., Interpretation of hyperspectral remote-sensing imagery by spectrum matching and look-up tables, Applied Optics 44 (17) (2005) 3576e3592. [70] J. Hedley, C. Roelfsema, S.R. Phinn, Efficient radiative transfer model inversion for remote sensing applications, Remote Sensing of Environment 113 (11) (2009) 2527e2532. [71] A. Albert, C.D. Mobley, An analytical model for subsurface irradiance and remote sensing reflectance in deep and shallow case-2 waters, Optics Express 11 (22) (2003) 2873e2890. [72] S. Maritorena, A. Morel, B. Gentili, Diffuse reflectance of oceanic shallow waters: influence of water depth and bottom albedo, Limnology and Oceanography 39 (7) (1994) 1689e1703. [73] A. Albert, G. Peter, Inversion of irradiance and remote sensing reflectance in shallow water between 400 and 800 nm for calculations of water and bottom properties, Applied Optics 45 (10) (2006) 2331e2343. [74] C. Giardino, G. Candiani, M. Bresciani, Z. Lee, S. Gagliano, M. Pepe, Bomber: a tool for estimating water quality and bottom properties from remote sensing images, Computers and Geosciences 45 (2012) 313e318. [75] S. Jay, M. Guillaume, Regularized estimation of bathymetry and water quality using hyperspectral remote sensing, International Journal of Remote Sensing 37 (2) (2016) 263e289. [76] W.M. Klonowski, P.R.C.S. Fearns, M.J. Lynch, Retrieving key benthic cover types and bathymetry from hyperspectral imagery, Journal of Applied Remote Sensing 1 (1) (2007) 011505. [77] Z. Lee, K.L. Carder, C.D. Mobley, R.G. Steward, J.S. Patch, Hyperspectral remote sensing for shallow waters: 2. deriving bottom depths and water properties by optimization, Applied Optics 38 (18) (1999) 3831e3843.

404 SECTION j III Application fields [78] Z.P. Lee, K.L. Carder, R.A. Arnone, Deriving inherent optical properties from water color: a multiband quasi-analytical algorithm for optically deep waters, Applied Optics 41 (27) (2002) 5755e5772. [79] P. Roland, La conservation des dunes littorales implique-t-elle leur stabilisation? l’exemple de la coˆte atlantique, Natures Sciences Socie´te´s 11 (3) (2003) 288e294. [80] K.E. Carpenter, M. Abrar, G. Aeby, R.B. Aronson, S. Banks, A. Bruckner, A. Chiriboga, J. Corte´s, J.C. Delbeek, L. DeVantier, et al., One-third of reef-building corals face elevated extinction risk from climate change and local impacts, Science 321 (5888) (2008) 560e563. [81] C.D. Clark, P.J. Mumby, J.R.M. Chisholm, J. Jaubert, S. Andrefouet, Spectral discrimination of coral mortality states following a severe bleaching event, International Journal of Remote Sensing 21 (11) (2000) 2321e2327. [82] T. Kutser, D.L.B. Jupp, On the possibility of mapping living corals to the species level based on their optical signatures, Estuarine, Coastal and Shelf Science 69 (3e4) (2006) 607e614. [83] P. Mouquet, T. Bajjouk, L. Maurel, A. Cebeillac, R. Le Goff, M. Ropert, Atlas des re´sultats du traitement des images hyperspectrales et des donne´es lidar sur les plateformes re´cifales de la re´union, 2014. [84] Hyscores, evaluation de l’e´tat de sante´ des coraux par imagerie hyperspectrale, 2015. http://sextant.ifremer.fr/fr/web/ocean_indien/hyscores. [85] T. Bajjouk, M. Pascal, M. Ropert, Q. Jean-Pascal, L. Hoarau, L. Bigot, N. Le Dantec, C. Delacourt, J. Populus, Detection of changes in shallow coral reefs status: towards a spatial approach using hyperspectral and multispectral data, Ecological Indicators 96 (2019) 174e191. [86] T. Petit, T. Bajjouk, M. Pascal, S. Rochette, V. Benoit, C. Delacourt, Hyperspectral remote sensing of coral reefs by semi-analytical model inversionecomparison of different inversion setups, Remote Sensing of Environment 190 (2017) 348e365. [87] J.D. Hedley, A.R. Harborne, P.J. Mumby, Simple and robust removal of sun glint for mapping shallow-water benthos, International Journal of Remote Sensing 26 (10) (2005) 2107e2112. [88] G. Sicot, M. Lennon, D. Corman, F. Gauthiez, Estimation of the sea bottom spectral reflectance in shallow water with hyperspectral data, in: 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), IEEE, 2015, pp. 2311e2314. [89] V. E Brando, J. M Anstee, M. Wettle, A.G. Dekker, S.R. Phinn, C. Roelfsema, A physics based retrieval and quality assessment of bathymetry from suboptimal hyperspectral data, Remote Sensing of Environment 113 (4) (2009) 755e770. [90] S. Jay, M. Guillaume, C. Malik, A. Minghelli, Y. Deville, B. Lafrance, V. Serfaty, Predicting minimum uncertainties in the inversion of ocean color geophysical parameters based on cramer-rao bounds, Optics Express 26 (2) (2018) A1eA18. [91] G.A. Shaw, H.K. Burke, Spectral imaging for remote sensing, Lincoln Laboratory Journal 14 (1) (2003) 3e28. [92] K.J. Bormann, R.D. Brown, C. Derksen, T.H. Painter, Estimating snow-cover trends from space, Nature Climate Change (2018) 1. [93] P. de Rosnay, G. Balsamo, C. Albergel, J. Mun˜oz-Sabater, L. Isaksen, Initialisation of land surface variables for numerical weather prediction, Surveys in Geophysics 35 (3) (2014) 607e621. [94] R.L. Armstrong, K. Rittger, M.J. Brodzik, A. Racoviteanu, A.P. Barrett, S.-J. Singh Khalsa, B. Raup, A.F. Hill, A.L. Khan, A.M. Wilson, et al., Runoff from glacier ice and seasonal

Applications in remote sensingdnatural landscapes Chapter j 3.1

[95]

[96] [97]

[98] [99]

[100]

[101]

[102] [103]

[104]

[105]

[106]

[107]

[108]

405

snow in high Asia: separating melt water sources in river flow, Regional Environmental Change (2018) 1e13. M. Reveillet, C. Vincent, D. Six, A. Rabatel, Sensitivity of surface mass balance based on direct measurements made on four distinct French alpine glaciers over the last two decades, and melt models performances comparison, in: AGU Fall Meeting Abstracts, 2015. E. Thibert, N. Eckert, C. Vincent, Climatic drivers of seasonal glacier mass balances: an analysis of 6 decades at Glacier de Sarennes (French Alps), The Cryosphere 7 (1) (2013) 47. V.V. Salomonson, I. Appel, Estimating fractional snow cover from MODIS using the normalized difference snow index, Remote Sensing of Environment 89 (3) (2004) 351e360. J. Dozier, Spectral signature of alpine snow cover from the landsat thematic mapper, Remote Sensing of Environment 28 (1989) 9e22. T.H. Painter, K. Rittger, C. McKenzie, P. Slaughter, R.E. Davis, J. Dozier, Retrieval of subpixel snow covered area, grain size, and albedo from MODIS, Remote Sensing of Environment 113 (4) (2009) 868e879. S. Pascal, R. Mathieu, Y. Arnaud, Subpixel monitoring of the seasonal snow cover with MODIS at 250 m spatial resolution in the southern alps of New Zealand: methodology and accuracy assessment, Remote Sensing of Environment 113 (1) (2009) 160e181. J.M. Bioucas-Dias, A. Plaza, D. Nicolas, M. Parente, Du Qian, G. Paul, J. Chanussot, Hyperspectral unmixing overview: geometrical, statistical, and sparse regression-based approaches, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 5 (2) (2012) 354e379. H. Bruce, Theory of Reflectance and Emittance Spectroscopy, Cambridge university press, 2012. S. Pascal, R. Mathieu, Y. Arnaud, M.M. Khan, J. Chanussot, Improving MODIS spatial resolution for snow mapping using wavelet fusion and arsis concept, IEEE Geoscience and Remote Sensing Letters 5 (1) (2008) 78e82. T. Ranchin, L. Wald, Fusion of high spatial and spectral resolution images: the arsis concept and its implementation, Photogrammetric Engineering and Remote Sensing 66 (1) (2000) 49e61. T. Masson, M. Dumont, M. Dalla Mura, S. Pascal, S. Gascoin, J.-P. Dedieu, J. Chanussot, An assessment of existing methodologies to retrieve snow cover fraction from MODIS data, Remote Sensing 10 (4) (2018) 619. L. Drumetz, M.-A. Veganzones, H. Simon, R. Phlypo, J. Chanussot, C. Jutten, Blind hyperspectral unmixing using an extended linear mixing model to address spectral variability, IEEE Transactions on Image Processing 25 (8) (2016) 3890e3905. D.P. Tittensor, M. Walpole, S.L.L. Hill, D.G. Boyce, G.L. Britten, N.D. Burgess, S.H.M. Butchart, P.W. Leadley, E.C. Regan, R. Alkemade, R. Baumung, C. Bellard, L. Bouwman, N.J. Bowles-Newark, A.M. Chenery, W.W.L. Cheung, V. Christensen, H.D. Cooper, A.R. Crowther, M.J.R. Dixon, A. Galli, V. Gaveau, R.D. Gregory, N.L. Gutierrez, T.L. Hirsch, R. Hoft, S.R. Januchowski-Hartley, M. Karmann, C.B. Krug, F.J. Leverington, J. Loh, R.K. Lojenga, K. Malsch, A. Marques, D.H.W. Morgan, P.J. Mumby, T. Newbold, K. Noonan-Mooney, S.N. Pagad, B.C. Parks, H.M. Pereira, T. Robertson, C. Rondinini, L. Santini, J.P.W. Scharlemann, S. Schindler, U.R. Sumaila, L.S.L. Teh, J. van Kolck, P. Visconti, Y. Ye, A mid-term analysis of progress toward international biodiversity targets, Science 346 (6206) (2014) 241e244. G. Ceballos, P.R. Ehrlich, A.D. Barnosky, A. Garcia, R.M. Pringle, T.M. Palmer, Accelerated modern human-induced species losses: Entering the sixth mass extinction, Science Advances 1 (5) (2015) e1400253.

406 SECTION j III Application fields [109] R.J. Morris, Anthropogenic impacts on tropical forest biodiversity: a network structure and ecosystem functioning perspective, Philosophical Transactions of the Royal Society B: Biological Sciences 365 (1558) (2010) 3709e3718. [110] F. Stuart Chapin III, E.S. Zavaleta, V.T. Eviner, R.L. Naylor, P.M. Vitousek, H.L. Reynolds, D.U. Hooper, S. Lavorel, O.E. Sala, S.E. Hobbie, M.C. Mack, S. Dı´az, Consequences of changing biodiversity, Nature 405 (6783) (2000) 234e242. [111] S. Dı´az, F. Joseph, F. Stuart Chapin, D. Tilman, Biodiversity loss threatens human wellbeing, PLoS Biology 4 (8) (2006) e277. [112] C. Secades, B. O’Connor, C. Brown, M. Walpole, UNEP World Conservation Monitoring Centre, Secretariat of the Convention on Biological Diversity, Earth Observation for Biodiversity Monitoring: a Review of Current Approaches and Future Opportunities for Tracking Progress Towards the Aichi Biodiversity Targets, 2014. [113] J. Walter, J. Cavender-Bares, P. Ryan, D. Schimel, F.W. Davis, G.P. Asner, R. Guralnick, J. Kattge, A.M. Latimer, M. Paul, M.E. Schaepman, M.P. Schildhauer, F.D. Schneider, F. Schrodt, U. Stahl, S.L. Ustin, Monitoring plant functional diversity from space, Nature Plants 2 (3) (2016) 16024. [114] H.M. Pereira, S. Ferrier, M. Walters, G.N. Geller, R.H.G. Jongman, R.J. Scholes, M.W. Bruford, N. Brummitt, S.H.M. Butchart, A.C. Cardoso, N.C. Coops, E. Dulloo, D.P. Faith, J. Freyhof, R.D. Gregory, C. Heip, R. Hoft, G. Hurtt, W. Jetz, D.S. Karp, M.A. McGeoch, D. Obura, Y. Onoda, N. Pettorelli, B. Reyers, R. Sayre, J.P.W. Scharlemann, S.N. Stuart, E. Turak, M. Walpole, M. Wegmann, Essential biodiversity variables, Science 339 (6117) (2013) 277e278. [115] A.K. Skidmore, N. Pettorelli, N.C. Coops, G.N. Geller, M. Hansen, R. Lucas, C.A. Mu¨cher, B. O’Connor, M. Paganini, H.M. Pereira, M.E. Schaepman, W. Turner, T. Wang, M. Wegmann, Environmental science: agree on biodiversity metrics to track from space, Nature 523 (7561) (2015) 403e405. [116] N. Pettorelli, M. Wegmann, A. Skidmore, S. Mu¨cher, T.P. Dawson, M. Fernandez, R. Lucas, M.E. Schaepman, T. Wang, B. O’Connor, H. Robert, G. Jongman, P. Kempeneers, S. Ruth, A.K. Leidner, M. Bo¨hm, K.S. He, H. Nagendra, G. Dubois, T. Fatoyinbo, M.C. Hansen, M. Paganini, M. Helen, de Klerk, G.P. Asner, J.T. Kerr, A.B. Estes, D.S. Schmeller, U. Heiden, D. Rocchini, H.M. Pereira, T. Eren, N. Fernandez, A. Lausch, M.A. Cho, D. Alcaraz-Segura, M.A. McGeoch, W. Turner, A. Mueller, V. StLouis, J. Penner, P. Vihervaara, A. Belward, B. Reyers, G.N. Geller, Framing the concept of satellite remote sensing essential biodiversity variables: challenges and future directions, Remote Sensing in Ecology and Conservation 2 (3) (2016) 122e131. [117] J. Vanden Borre, D. Paelinckx, C.A. Mu¨cher, L. Kooistra, B. Haest, G. De Blust, M. Anne, Schmidt, Integrating remote sensing in Natura 2000 habitat monitoring: prospects on the way forward, Journal for Nature Conservation 19 (2) (2011) 116e125. [118] A. Townsend, G. Asner, C. Cleveland, The biogeochemical heterogeneity of tropical forests, Trends in Ecology and Evolution 23 (8) (2008) 424e431. ¨ . Niinemets, L. Poorter, I.J. Wright, R. Villar, Causes and consequences of [119] H. Poorter, U variation in leaf mass per area (LMA): a meta-analysis: Tansley review, New Phytologist 182 (3) (2009) 565e588. [120] S. Diaz, M. Cabido, F. Casanoves, Plant functional traits and environmental filters at a regional scale, Journal of Vegetation Science 9 (1) (1998) 113e122. [121] S. Lavorel, E. Garnier, Predicting changes in community composition and ecosystem functioning from plant traits: revisiting the Holy Grail, Functional Ecology 16 (5) (2002) 545e556.

Applications in remote sensingdnatural landscapes Chapter j 3.1

407

[122] G.P. Asner, R.E. Martin, Spectranomics: emerging science and conservation opportunities at the interface of biodiversity and remote sensing, Global Ecology and Conservation 8 (2016) 212e219. [123] G.P. Asner, R.E. Martin, R. Tupayachi, R. Emerson, P. Martinez, F. Sinca, V. George, N. Powell, S.J. Wright, A.E. Lugo, Taxonomy and remote sensing of leaf mass per area (LMA) in humid tropical forests, Ecological Applications 21 (1) (2011) 85e98. [124] G. le Maire, C. Franc¸ois, E. Dufreˆne, Towards universal broad leaf chlorophyll indices using PROSPECT simulated database and hyperspectral reflectance measurements, Remote Sensing of Environment 89 (1) (2004) 1e28. [125] A.A. Gitelson, G.P. Keydan, M.N. Merzlyak, Three-band model for noninvasive estimation of chlorophyll, carotenoids, and anthocyanin contents in higher plant leaves, Geophysical Research Letters 33 (11) (2006) L11402. [126] S. Jacquemoud, J. Verdebout, G. Schmuck, G. Andreoli, B. Hosgood, Investigation of leaf biochemistry by statistics, Remote Sensing of Environment 54 (3) (1995) 180e188. [127] J.-B. Fe´ret, A.A. Gitelson, S.D. Noble, S. Jacquemoud, PROSPECT-D: towards modeling leaf optical properties through a complete lifecycle, Remote Sensing of Environment 193 (2017) 204e215. [128] J.-B. Fe´ret, G. le Maire, S. Jay, D. Berveiller, R. Bendoula, G. Hmimina, A. Cheraiet, J.C. Oliveira, F.J. Ponzoni, T. Solanki, F. de Boissieu, J. Chave, Y. Nouvellon, A. PorcarCastell, C. Proisy, K. Soudani, J.-P. Gastellu-Etchegorry, M.-J. Lefe`vre-Fonollosa, Estimating leaf mass per area and equivalent water thickness based on leaf optical properties: potential and limitations of physical modeling and machine learning, Remote Sensing of Environment (2018) (in press), https://doi.org/10.1016/j.rse.2018.11.002. [129] S. Jacquemoud, S.L. Ustin, J. Verdebout, G. Schmuck, G. Andreoli, B. Hosgood, Estimating leaf biochemistry using the PROSPECT leaf optical properties model, Remote Sensing of Environment 56 (3) (1996) 194e202. [130] J.-B. Fe´ret, C. Franc¸ois, A. Gitelson, G.P. Asner, K.M. Barry, C. Panigada, A.D. Richardson, S. Jacquemoud, Optimizing spectral indices and chemometric analysis of leaf chemical properties using radiative transfer modeling, Remote Sensing of Environment 115 (10) (2011) 2742e2750. [131] G.P. Asner, R.E. Martin, Airborne spectranomics: mapping canopy chemical and taxonomic diversity in tropical forests, Frontiers in Ecology and the Environment 7 (5) (2009) 269e276. [132] J.-B. Fe´ret, G.P. Asner, Spectroscopic classification of tropical forest species using radiative transfer modeling, Remote Sensing of Environment 115 (9) (2011) 2415e2422. [133] J.-B. Fe´ret, G.P. Asner, Tree species discrimination in tropical forests using airborne imaging spectroscopy, IEEE Transactions on Geoscience and Remote Sensing 51 (1) (2013) 73e84. [134] C.A. Baldeck, G.P. Asner, R.E. Martin, C.B. Anderson, D.E. Knapp, J.R. Kellner, S.J. Wright, Operational tree species mapping in a diverse tropical forest with airborne imaging spectroscopy, PLoS One 10 (7) (2015) e0118403. [135] J. Cavender-Bares, J. Meireles, J. Couture, M. Kaproth, C. Kingdon, A. Singh, S. Serbin, A. Center, E. Zuniga, P. George, P. Townsend, Associations of leaf spectra with genetic and phylogenetic variation in Oaks: prospects for remote detection of biodiversity, Remote Sensing 8 (3) (2016) 221. [136] C. Chavana-Bryant, Y. Malhi, J. Wu, G.P. Asner, A. Anastasiou, B.J. Enquist, E.G.C. Caravasi, C.E. Doughty, S.R. Saleska, R.E. Martin, F.F. Gerard, Leaf aging of

408 SECTION j III Application fields

[137]

[138]

[139]

[140]

[141]

[142] [143]

[144]

[145]

[146]

[147] [148]

[149]

[150] [151]

Amazonian canopy trees as revealed by spectral and physiochemical measurements, New Phytologist 214 (3) (2017) 1049e1063. J. Zhang, B. Rivard, A. Sa´nchez-Azofeifa, K. Castro-Esau, Intra- and inter-class spectral variability of tropical tree species at La Selva, Costa Rica: implications for species identification using HYDICE imagery, Remote Sensing of Environment 105 (2) (2006) 129e141. K.L. Castro-Esau, G.A. Sanchez-Azofeifa, B. Rivard, S.J. Wright, M. Quesada, Variability in leaf optical properties of Mesoamerican trees and the potential for species classification, American Journal of Botany 93 (4) (2006) 517e530. H. Feilhauer, T. Schmid, U. Faude, S. Sa´nchez-Carrillo, C. Santos, Are remotely sensed traits suitable for ecological analysis? A case study of long-term drought effects on leaf mass per area of wetland vegetation, Ecological Indicators 88 (2018) 232e240. M.P. Ferreira, J.-B. Fe´ret, E. Grau, J.-P. Gastellu-Etchegorry, C.H. do Amaral, Y.E. Shimabukuro, C.R. de Souza Filho, Retrieving structural and chemical properties of individual tree crowns in a highly diverse tropical forest with 3D radiative transfer modeling and imaging spectroscopy, Remote Sensing of Environment 211 (2018) 276e291. A. Singh, S.P. Serbin, B.E. McNeil, C.C. Kingdon, P.A. Townsend, Imaging spectroscopy algorithms for mapping canopy foliar chemical and morphological traits and their uncertainties, Ecological Applications 25 (8) (2015) 2180e2197. K. Chadwick, G. Asner, Organismic-scale remote sensing of canopy foliar traits in lowland tropical forests, Remote Sensing 8 (2) (2016) 87. G.P. Asner, R.E. Martin, C.B. Anderson, D.E. Knapp, Quantifying forest canopy traits: imaging spectroscopy versus field survey, Remote Sensing of Environment 158 (2015) 15e27. G.P. Asner, S.L. Ustin, P.A. Townsend, R.E. Martin, K.D. Chadwick, Forest biophysical and biochemical properties from hyperspectral and LiDAR remote sensing, in: Land Resources Monitoring, Modeling, and Mapping with Remote Sensing, Remote Sensing Handbook, CRC Press, 2015, pp. 429e448. R. Martin, K. Chadwick, P. Brodrick, L. Carranza-Jimenez, N. Vaughn, G. Asner, An approach for foliar trait retrieval from airborne imaging spectroscopy of tropical forests, Remote Sensing 10 (2) (2018) 199. G.P. Asner, P.G. Brodrick, C.B. Anderson, N. Vaughn, D.E. Knapp, R.E. Martin, Progressive forest canopy water loss during the 2012e2015 California drought, Proceedings of the National Academy of Sciences 113 (2) (2016) E249eE255. S.L. Ustin, J.A. Gamon, Remote sensing of plant functional types, New Phytologist 186 (4) (2010) 795e816. H. Feilhauer, B. Somers, S. van der Linden, Optical trait indicators for remote sensing of plant species composition: predictive power and seasonal variability, Ecological Indicators 73 (2017) 825e833. M.W. Palmer, T. Wohlgemuth, P. Earls, J.R. Are´valo, S.D. Thompson, Opportunities for long-term ecological research at the Tallgrass Prairie Preserve, Oklahoma, in: K. Lajtha, K. Vanderbilt (Eds.), Proceedings of ILTER Regional Workshop, 2000, pp. 123e128. Budapest, Hungary. D. Rocchini, A. Chiarucci, A. Steven, Loiselle, Testing the spectral variation hypothesis by using satellite multispectral images, Acta Oecologica 26 (2) (2004) 117e120. J.-B. Fe´ret, G.P. Asner, Mapping tropical forest canopy diversity using high-fidelity imaging spectroscopy, Ecological Applications 24 (6) (2014) 1289e1296.

Applications in remote sensingdnatural landscapes Chapter j 3.1

409

[152] C. Baldeck, G. Asner, Estimating vegetation beta diversity from airborne imaging spectroscopy and unsupervised clustering, Remote Sensing 5 (5) (2013) 2057e2071. [153] C.A. Baldeck, M.S. Colgan, J.-B. Fe´ret, S.R. Levick, R.E. Martin, G.P. Asner, Landscapescale variation in plant community composition of an African savanna from airborne species mapping, Ecological Applications 24 (1) (2014) 84e93. [154] H. Feilhauer, U. Faude, S. Schmidtlein, Combining Isomap ordination and imaging spectroscopy to map continuous floristic gradients in a heterogeneous landscape, Remote Sensing of Environment 115 (10) (2011) 2513e2524. [155] A.K. Schweiger, J. Cavender-Bares, P.A. Townsend, S.E. Hobbie, M.D. Madritch, R. Wang, D. Tilman, J.A. Gamon, Plant spectral diversity integrates functional and phylogenetic components of biodiversity and predicts ecosystem function, Nature Ecology and Evolution 2 (6) (2018) 976e982. [156] R. Wang, J.A. Gamon, J. Cavender-Bares, P.A. Townsend, A.I. Zygielbaum, The spatial sensitivity of the spectral diversity-biodiversity relationship: an experimental test in a prairie grassland, Ecological Applications 28 (2) (2018) 541e556. [157] R.H. Whittaker, Evolution and measurement of species diversity, Taxon 21 (2/3) (1972) 213. [158] T. Hanna, K. Ruokolainen, Analyzing or explaining beta diversity? Understanding the targets of different methods of analysis, Ecology 87 (11) (2006) 2697e2708. [159] G.P. Asner, D.E. Knapp, B. Joseph, R.O. Green, T. Kennedy-Bowdoin, M. Eastwood, R.E. Martin, C. Anderson, C.B. Field, Carnegie Airborne Observatory-2: increasing science data dimensionality via high-fidelity multi-sensor fusion, Remote Sensing of Environment 124 (2012) 454e465. [160] M.D. Madritch, C.C. Kingdon, A. Singh, K.E. Mock, R.L. Lindroth, P.A. Townsend, Imaging spectroscopy links aspen genotype with below-ground processes at landscape scales, Philosophical Transactions of the Royal Society B: Biological Sciences 369 (1643) (2014) 20130194. [161] I.J. Wright, P.B. Reich, M. Westoby, D.D. Ackerly, Z. Baruch, F. Bongers, J. CavenderBares, T. Chapin, J.H.C. Cornelissen, M. Diemer, J. Flexas, E. Garnier, P.K. Groom, J. Gulias, K. Hikosaka, B.B. Lamont, T. Lee, W. Lee, C. Lusk, J.J. Midgley, M.-L. Navas, ¨ . Niinemets, J. Oleksyn, N. Osada, H. Poorter, P. Poot, L. Prior, V.I. Pyankov, C. Roumet, U S.C. Thomas, M.G. Tjoelker, E.J. Veneklaas, R. Villar, The worldwide leaf economics spectrum, Nature 428 (6985) (2004) 821e827. [162] J.L.D. Osnas, J.W. Lichstein, P.B. Reich, S.W. Pacala, Global leaf trait relationships: mass, area, and the leaf economics spectrum, Science 340 (6133) (2013) 741e744. [163] J.-P. Gastellu-Etchegorry, T. Yin, N. Lauret, T. Cajgfinger, T. Gregoire, E. Grau, J.B. Fe´ret, M. Lopes, G. Jordan, G. Dedieu, Z. Malenovsky´, B. Cook, D. Morton, J. Rubio, S. Durrieu, G. Cazanave, E. Martin, T. Ristorcelli, Discrete anisotropic radiative transfer (DART 5) for modeling airborne and satellite spectroradiometer and LIDAR acquisitions of natural and urban landscapes, Remote Sensing 7 (2) (2015) 1667e1701. [164] G. Vincent, C. Antin, M. Laurans, J. Heurtebize, S. Durrieu, C. Lavalley, J. Dauzat, Mapping plant area index of tropical evergreen forest by airborne laser scanning. A crossvalidation study using LAI2200 optical sensor, Remote Sensing of Environment 198 (2017) 254e266. [165] L. Candela, R. Formaro, R. Guarini, R. Loizzo, F. Longo, G. Varacalli, The PRISMA mission, in: 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), IEEE, Beijing, China, 2016, pp. 253e256. [166] T. Matsunaga, A. Iwasaki, S. Tsuchida, K. Iwao, J. Tanii, O. Kashimura, R. Nakamura, H. Yamamoto, S. Kato, K. Obata, K. Mouri, T. Tachikawa, Current status of hyperspectral

410 SECTION j III Application fields imager suite (HISUI) onboard international space station (ISS), in: 2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), IEEE, Fort Worth, TX, 2017, pp. 443e446. [167] L. Pedro, M. Schwieder, S. Suess, A. Okujeni, L. Galva˜o, S. Linden, P. Hostert, Monitoring natural ecosystem and ecological gradients: perspectives with EnMAP, Remote Sensing 7 (10) (2015) 13098e13119.