Development and testing of a hyperspectral camera network to monitor plant canopies and sediment transport dynamics

Abstract

There is growing interest in hyperspectral imaging to complement observation needs and techniques required to capture the critical zone dynamics. It is already widely used in remote sensing satellite imagery, for regional-scale monitoring of canopies (Asner et al. 2004), or suspended sediment transport (Yepez et al. 2017). Spectral imaging offers dense, remote and non-intrusive measurement coverage. Its implementation at fixed-station for fine temporal monitoring would ensure maximum temporal coverage to study the phenology and functioning of ecosystems (vegetation-water-soil interactions) and watersheds (sediment dynamics), at integrative scales (e.g. watershed outlets), or over experimental plots. On-site hyperspectral data also enable links to the regional scale through cross-comparison with data from space (de Moura et al. 2017). It would then enables to better control measurement biases, offering in that way better opportunity for standardizing observables, as required by international research infrastructures. In recent years, both technological progresses and applications for commercial uses made these kind of cameras more reliable, compact, and affordable, making feasible on-site hand-held or UAV-based experiments (Stuart et al. 2019). Nevertheless, deployment for continuous monitoring remains uncommon, and limited to specific applications (de Moura et al. 2017, Woodgate et al. 2020), due to a still high instrumental complexity and costs. Furthermore, correct data exploitation requires a complete mastery of the calibration, acquisition, normalization and processing chain, that can be complex with “black-box” commercial systems. The development of a dedicated spectral camera is thus preferred. Such a camera is developed within the program TERRA FORMA from the French Agency for Research (Longuevergne et al. 2022). This program aims to implement integrated socio-ecosystem observatories, in support of the French RZA and OZCAR infrastructures, by developing and deploying a dozen types of state-of-the-art sensors dedicated to environmental monitoring at national-scale until 2029. A part of this project is dedicated to the deployment up to 20 spectral cameras, within two scientific topics: monitoring of plant canopies, monitoring of suspended sediment dynamics in rivers. monitoring of plant canopies, monitoring of suspended sediment dynamics in rivers. The instrumental solution we are implementing is based on developments carried out at IPAG since 2016 in compact spectral imaging for spaceborne Earth Observation (Gousset et al. 2019, Le Coarer et al. 2021). In addition to its compactness and optical simplicity, the main advantage of this kind of camera lies in its ability to acquire all spectral and spatial information in a single acquisition (“snapshot”) of a fraction of a second. By opposite to pushbroom or linescanner concepts, which require tens of seconds of exposure under stable illumination conditions. The TERRA FORMA camera complements these instruments with a frugal, less expensive solution, suitable for deployment as a stand-alone fixed station or for handle-held/UAV acquisitions on the field. Since May 2024, we integrated and tested in laboratory an operational camera (Fig. 1), with the following specifications: Field of view 22 by 12°, for 365 by 200 pixels 1 cm / pixel at 9 m distance 42 spectral channels between 400 and 780 nm (up to 850) Spectral resolution 10 nm (up to 6 nm) 10 x 10 x 6 cm, 0.6 kg, powered by LiPo battery Field of view 22 by 12°, for 365 by 200 pixels 1 cm / pixel at 9 m distance 42 spectral channels between 400 and 780 nm (up to 850) Spectral resolution 10 nm (up to 6 nm) 10 x 10 x 6 cm, 0.6 kg, powered by LiPo battery We carried out a first field test in August 2024 at the eLTER site Lautaret / Roche Noire (French Alps). During this single day of acquisition, we acquired data over the landscape jointly to a reference commercial non-imaging spectrometer. This last is shown on Fig. 2, demonstrating a good adequacy between hyperspectral data from the camera and reference spectra. The next steps for 2025 are on site campaigns, lasting 3 to 6 months at fixed stations on pilot sites. On the biodiversity topic: acquisition during a full growing season in a snow-covered mountain grassland equipped with a flux tower should enable: To compare the series of data from hyperspectral imagery with the installed multi-spectral NDVI sensor (only two channels in red and near infrared). To compare spectral measurements with balances of radiative fluxes, and with CO 2 and H 2 O exchanges in the soil-plant-atmosphere continuum. To identify the best optical proxies for inferring vegetation water status and CO 2 fixation capacity during a season. To compare the series of data from hyperspectral imagery with the installed multi-spectral NDVI sensor (only two channels in red and near infrared). To compare spectral measurements with balances of radiative fluxes, and with CO 2 and H 2 O exchanges in the soil-plant-atmosphere continuum. To identify the best optical proxies for inferring vegetation water status and CO 2 fixation capacity during a season. Mid-term objective is to be able to increase the effective footprint of the tower, then to be able to infer canopy function and structure using imagery, through integrated and continuous measurement of several biodiversity parameters at the same time, complementary to data collected as part of the eLTER and ICOS infrastructures. On the hydrology topic: another camera will be deployed on hydrological stations (campus of Grenoble, then Galabre river (Legout et al. 2021)). The aggregation of data should enable: To identify optical proxies for quantifying suspended solids concentrations. To evaluate the robustness of this approach in a concentration range from 0 to a few tens of g/l, currently well measured by the combined turbidimetry and sampling approach (Navratil et al. 2011). To identify optical proxies capable of discriminating between the different types of suspended solids transported in rivers during floods. To apply an approach based on these optical proxies to trace the sources of suspended solids using mixture models, and compare these results with those obtained using the spectro-colorimetric manual suspended solids tracing method implemented on the Galabre site since 2013 (Legout et al. 2013). To identify optical proxies for quantifying suspended solids concentrations. To evaluate the robustness of this approach in a concentration range from 0 to a few tens of g/l, currently well measured by the combined turbidimetry and sampling approach (Navratil et al. 2011). To identify optical proxies capable of discriminating between the different types of suspended solids transported in rivers during floods. To apply an approach based on these optical proxies to trace the sources of suspended solids using mixture models, and compare these results with those obtained using the spectro-colorimetric manual suspended solids tracing method implemented on the Galabre site since 2013 (Legout et al. 2013). The final objective is to be able to complement in situ techniques (turbidimetry) and river sampling with a remote, robotized measurement method, providing better temporal coverage of flood episodes, more reliable than submerged sensors.

Plant phenotyping relevance

植物キャノピーの状態・機能を測定する専用ハイパースペクトルカメラを開発・試験しており、植物フェノタイピング用の取得基盤が研究の中心である。

Code and data availability

This conference abstract describes a hyperspectral camera (TERRA FORMA) and a single field test at Lautaret, but contains no data availability statement, no public repository deposit, no author code/model URL, and no supplement. The hyperspectral acquisitions and spectra shown in figures are not stated as publicly可用, 0

No evidence-backed public reproduction asset is currently recorded.