Can we trust multispectral drone datasets for eLTER variables monitoring? Sensors sensitivity and applications.

Multispectral drone data are widely used for monitoring vegetated critical zones. Numerous indices derived from remotely sensed data provide objective and spatially comprehensive observations of land surfaces. These indices are employed in various applications, including precision agriculture (Deng et al. 2018), water stress detection in viticulture (Santesteban et al. 2017; Kandylakis et al. 2020), forest ecology for phenology and health assessment (Ecke et al. 2024, Fraser and Congalton 2021), ecosystem habitat mapping (Alvarez-Vanhard et al. 2020), and soil moisture monitoring (Bertalan et al. 2022). In all these cases, vegetation indices remain fundamental tools for vegetation monitoring. Most studies rely on reflectance data from multispectral sensors calibrated using manufacturer-provided reference panels. These 'plug-and-play' solutions facilitate widespread applications, yet reflectance values are rarely validated against ground-truth spectral signatures. Consequently, users often place unquestioning trust in pre-processed data. Additionally, spectral bands vary across different sensors, much like satellite sensors, further complicating data consistency. This study presents the results of a calibration and validation experiment. Four multispectral sensors (Parrot Sequoia, Micasense Dual MX and Altum-PT, and Spectral Device) with distinct spectral characteristics (spatial and spectral resolution, bandwidth) were compared with spectral signatures obtained using an ASD Fieldspec Pro spectroradiometer. Spectral data were collected along transects in the Sougéal marsh (N 48°51’, W -1°50’, France) , which features diverse herbaceous habitats and varying soil moisture conditions (Fig. 1). The comparison highlights sensor sensitivity and their capacity to provide consistent data. Preliminary results indicate significant quality differences between sensors. Two applications were conducted. The first analyzed NDVI derived from NIR and red bands to evaluate how spectral characteristics and sensitivity impact this commonly used vegetation index. The second application focused on soil moisture monitoring. One sensor, equipped with SWIR bands, was used to estimate soil moisture using the OPTRAM model (Sadeghi et al. 2017). Index values were compared against in situ spectral signatures and soil moisture measurements obtained with a Tetra Probe.