The US National Academies Decadal Survey calls out the need for global imaging spectroscopy to characterize “functional traits of terrestrial plants (structure, physiology, phenology, reproduction, and biochemistry)” both to “determine the patterns of energy, carbon, water and nutrient fluxes for terrestrial ecosystems” and “provide a direct, mechanistic link to biological diversity.” These observables are characterized as either “most” or “very” important (questions E-1 and E-3 in Decadal Survey). Numerous airborne campaigns and some work from satellites such as EO-1 Hyperion demonstrate reliable measurements (<25% uncertainty) for a suite of 20-30 foliar functional traits from imaging spectroscopy. However, with the increasing availability of airborne data (e.g. from APEX, NEON and AVIRIS-NG, HySpex, HyMap as well as UAS campaigns) and numerous forthcoming spaceborne missions, there is now an urgent need to harmonise approaches to foliar trait mapping across platforms to facilitate rigorous validation and data integration. This includes need for 1) networks of ground observations to enable comparative studies and cross-calibration, 2) agreement on sampling, field/lab measurement and scaling techniques, and 3) comparison and cooperation on testing retrieval methods that balance complexity with need for efficient processing of the vast volumes of forthcoming data. As well, the community needs to identify standards for computation of diversity metrics. Here we lay out the current status, as illustrated by campaigns in the US (NEON and HyspIRI preparatory efforts), Europe and India.

Although imaging spectroscopy has the potential to estimate quantitatively plant functional traits at regional-to-global scales, a large volume of the data and its informational redundancy increase operational demands for the data manipulation and processing. In 2018, we concluded the ESA study “Hyperspectral analysis and heterogeneous surface modelling”, which developed a method selecting appropriate spectral bands for a machine learning-based retrieval of biophysical plant traits. We utilized the Discrete Anisotropic Radiative Transfer (DART) to model a spectral database (look-up-table: LUT) of several thousand top-of-canopy reflectance signatures (spectral range of 400-2500 nm and spectral resolution of 1-2 nm) for maize, citrus orchard, spruce, beech and mixed floodplain forest study sites. Using LUT, we determined an effective method selecting optimal number and wavelengths of spectral bands to estimate chlorophyll a+b (Cab) and total carotenoid contents (Car), leaf water content (Cw), and leaf area index (LAI) with acceptable accuracy. The method, combining the Least Absolute Shrinkage and Selection Operator with the Sequential Feature Selection, was compared with reference method using all bands through Support Vector Regression (SVR) retrievals applied on DART simulations and airborne images. Both approaches yielded similar spatial patterns and ranges of the estimates. The most successful was Cw retrieval, while retrievals of Car and Cab were moderately good (overestimated by 2.5 and by 20 µg/cm², respectively). LAI retrieval was the least successful. We concluded that estimations from non-redundant spectral bands achieved a lower but acceptable accuracy. However, the band selection required significant time for a proper optimization of SVR predictive models.

An unprecedented spectroscopic data stream will soon become available with forthcoming Earth-observing satellite missions equipped with imaging spectroradiometers. This data stream will open up a vast array of opportunities to quantify a diversity of biochemical and structural vegetation properties. The processing requirements for such large data streams require reliable retrieval techniques enabling the spatiotemporally explicit quantification of biophysical variables.
Overall, retrieval methods can be categorized into: (1) parametric regression, including vegetation indices, shape indices and spectral transformations; (2) nonparametric regression, including linear and nonlinear machine learning regression algorithms; (3) physically based, including inversion of radiative transfer models (RTMs) using numerical optimization and look-up table approaches; and (4) hybrid regression methods, which combine RTM simulations with machine learning regression methods.
An overview of these categories with application to mapping vegetation properties will be given. In view of processing hyperspectral data, a critical aspect involves the challenge of dealing with spectral multicollinearity. The ability to provide robust estimates, retrieval uncertainties and acceptable retrieval processing speed are other important aspects in view of operational processing. Based on a qualitative evaluation of the strengths and weaknesses of the retrieval methods, recommendations will be given regarding their usage for operational processing of data streams coming from new-generation imaging spectrometer missions such as FLEX and CHIME. Putting these recommendations into practice, this contribution will provide examples of developed retrieval algorithms for inferring vegetation variables from FLEX and CHIME reflectance data. It will also discuss the latest status of the developed retrieval algorithms and the lessons learned.

Soil spectroscopy based on laboratory, field, and airborne data has been shown to be a proven method for the quantitative prediction of key soil properties in local areas when bare soils are exposed at the surface at appropriate and specific conditions, and ground data are available. With the recent and upcoming launches of the next generation of imaging spectroscopy satellite sensors in next years, a high potential to meet the demand for global top soil mapping and monitoring is appearing.
Nevertheless, the full maturity of imaging spectroscopy for soil mapping has not been reached due to some limitations in measurement scheme, data and software availability, soil spectral modelling approaches, and in the availability of appropriate global soil spectral libraries as well as global ground databases for calibration and validation. In this paper, we review the achievements and perspectives in soil mapping and monitoring from imaging spectroscopy. Digital soil mapping studies using airborne and simulated imaging spectroscopy data from several test sites will be shown for the quantitative prediction of key top-soil properties such as soil organic carbon, clay content, and soil degradation stages. Furthermore, new methodological advances have allowed to integrate operational algorithms or harmonized soil spectral libraries for enhanced soil surface mapping. Finally, current challenges, gaps and new directions toward a better exploitation of space-based imaging spectrometers for operational soil mapping are discussed, in particular, further developments at the international level in the direction of harmonization of data products, retrieval algorithms, and global databases are needed.

Agricultural soils are one of the most relevant topics in agriculture, as soil fertility is directly related to factors such as nutrients concentrations or availability, organic matter content, acidity, moisture, among others.
The objective of this communication is to present the preliminary results of a work in progress that aims to assess the value of novel and forthcoming hyperspectral satellite missions, such as ASI PRISMA, DLR EnMAP, and ESA Copernicus CHIME, in terms of their potential in soil properties retrieval.
This is exploited in the framework of the Grosseto (Italy) 2018 trial in which data were acquired by the airborne imaging spectrometer AVIRIS-NG contemporary to topsoil ground measurements. Ordinary kriging was applied to ground-collected data to obtain the spatial distribution of the soil properties and of their uncertainty. Spectral features analysis, based on band ratios or other formulation, and two different multivariate methods were compared for the calibration and validation of clay and soil organic carbon (SOC) data estimation models. In particular: i) literature spectral indexes, ii) partial least square regression (PLSR) and iii) random forest (RF) were applied.
Different methods were applied to both airborne original and spatially/spectrally resampled data in order to simulate the retrieval performances of a next-generation hyperspectral satellite mission like CHIME.
The preliminary results show the advantages of the high spatial and spectral resolutions of the available airborne dataset that were used for the retrieval of topsoil properties of agricultural soils in the investigated area.