Evaluation of phenospectral dynamics with Sentinel-2A using a bottom-up approach in a northern ombrotrophic peatland

• Published in: Remote sensing of environment 2018-10, Vol.216, p.544-560
• Main Authors: Arroyo-Mora, J.P., Kalacska, M., Soffer, R., Ifimov, G., Leblanc, G., Schaaf, E.S., Lucanus, O.
• Format: Article
• Language: English
• Subjects: Airborne hyperspectral; Airborne sensing; Bog; Carbon cycle; Carbon sinks; CASI; Change detection; Climate change; Data acquisition; Datasets; Ecosystems; Environmental changes; Greening; Growing season; Land product validation; Landscape; Mer Bleue; Mosaics; Multi-datasets; Peatlands; Phenological changes; Phenology; Phenospectral; Reflectance; Remote sensing; Satellites; Sentinel-2 MSI; Spatial data; Spectra; Spectral resolution; Spectral sensitivity; Spectroscopy; Spectrum analysis; Vegetation; Vegetation changes
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2018.07.021

Peatlands cover very large extents in northern regions and play a significant role in the global carbon cycle by functioning as a carbon sink. Large-scale satellite based monitoring systems, such as the Sentinel-2 Multispectral Instrument (MSI), are necessary to improve our understanding of how these ecosystems respond to climate change by providing verifiable land products. For instance, satellite-based land product validation approaches can benefit from airborne hyperspectral imagery and in-situ data, which provide higher spatial and spectral resolution baselines, ideal for measuring vegetation changes (e.g. phenology, LAI) at local scales. Here, we assessed the short-term phenospectral dynamics (spectral changes indicated by specific spectral features as a function of phenology) of five ombrotrophic peatland vegetation physiognomies over four dates at the Mer Bleue bog in Canada. We took advantage of a unique remote sensing data acquisition campaign aiming to validate Sentinel-2A land products, and analyzed three spatially and spectrally distinctive datasets (i.e. field spectra, VISNIR airborne hyperspectral imagery (HSI) and Sentinel-2A imagery) over the first half of the 2016 growing season. By implementing a bottom-up approach, first we assessed the airborne HSI's capability to detect phenological changes as compared to in-situ acquired field spectroscopy measurements in a 10 ha area at Mer Bleue and evaluated the spectral features characteristic of these phenological changes. Second, over the entire Mer Bleue area (28,000 ha), we compared a series of four Sentinel-2A images to four airborne HSI mosaics (spatially and spectrally resampled to Sentinel-2A) to assess the utility of Sentinel-2A for detecting small spectral variations due to phenological changes (i.e. greening). In addition, for this second comparison, three spectral vegetation indices were derived from the Sentinel-2A images and the airborne HSI mosaics. The spectral comparisons between the airborne HSI and the field spectroscopy data revealed clear phenological changes from the airborne HSI. For instance, a closer agreement between reflectance measured by the field spectrometer and the airborne HSI spectral response was found in the visible region (450–680 nm). A greater difference however, was consistently seen in the near-infrared region (681–866 nm) across the four dates. Narrow spectral features in three regions of the visible range (global minima, red absorption, green peak), indic