Modelling the Spatial Distribution of CO2 Fluxes in a Subalpine Grassland Plateau of the Italian Alps Using Multiple Airborne AISA Eagle Hyperspectral Sensor Observations and Sentinel-2 Simulated Data.

Abstract:

ESA’s satellite Sentinel-2 provides images of high spatial, spectral and temporal resolution, providing a high potential for biophysical parameter characteristics monitoring and for products validation at the Eddy Covariance (EC) towers. A set of 13 spectral bands is available ranging from the visible and NIR to SWIR, featuring four bands at 10 m, six bands at 20 m and three bands at 60 m spatial resolution. Depending on the presence of clouds, satellite data will be available every 10-15 days.

In comparison to the last sensors, Sentinel-2 incorporates three new spectral bands in the red-edge region which are particularly important for the retrieval and monitoring of biophysical parameter characteristics of dynamic ecosystems such as crops and grasslands.

Under the umbrella of the COST Action ES0903, a hyperspectral flight campaign was organised at Viote del Monte Bondone (Trento, Italy) and five EC towers were installed on subalpine grasslands characterised by extreme variability of ecosystem structural parameters. The aim of the campaign was to compare the performance of different vegetation indices (simulating Sentinel-2 bands) in estimating NEE of these grassland ecosystems and to explore the structural and radiation controls on CO₂ fluxes. The predictive capacity of partial least squares regression (PLSR) models using the full range of spectral data from different sites and from different airborne acquisitions was also assessed, and the appropriateness of the Sentinel-2 spatial resolution for ground-based flux upscaling was tested.

The high structural heterogeneity of the investigated canopies resulted in high NEE fluxes spatial variability within the 5 investigated towers, indicating that, within the same grassland vegetation type, there is an evident control of structural traits on photosynthesis. Including PAR into the model resulted in a general increase in the performance of the linear regression. Also, the high predictive capacity of PLSR models confirmed the potential of using the full spectrum in the models.

Our findings have major implications for upscaling terrestrial CO₂ fluxes to larger regions and call for more cross-site synthesis studies linking satellite observations, ground-based observations and ecosystem-scale CO₂ fluxes.