Mapping Alpine pasture biomass and nutritional value from hyperspectral field and spaceborne sensors

Pastures are important components of Alpine cultural landscapes. However, due to unbalanced grazing pressure and climate change, the quality of pastures is degrading, diminishing their capacity to deliver essential ecosystem services and leading to widespread abandonment. Adaptive grazing management practices can address the challenge of under- or over-exploitation of pasture resources, balancing livestock forage demand and pasture productivity. This study evaluates the capabilities of spaceborne hyperspectral data for mapping pasture productivity and nutritional characteristics (i.e. biomass, protein and fiber content) in mid- to high-elevation Alpine environments. Agronomic and reflectance data (350-2500 nm) were collected in the field across different vegetation phenological stages for 128 pasture samples in two distinct Alpine regions. Field spectra were convoluted to simulate satellite PRISMA hyperspectral and Sentinel-2 multispectral sensor configurations, both used to calibrate and validate Gaussian Process Regression machine learning models. Hyperspectral-based models significantly outperformed multispectral-based ones, with an average Rcv2 improvement of 58.05% and an average RMSEcv reduction of 9.83% (Rcv2: 0.67, RMSEcv: 225.79 gm−2 for AGB; Rcv2: 0.55, RMSEcv: 1.87 % for CP; Rcv2: 0.65, RMSEcv: 1.71 % for ADF; Rcv2: 0.62, RMSEcv: 2.66 % for NDF). The best-performing models were demonstrated on a real PRISMA scene, and the estimations were evaluated against in-situ data, showing overall good agreement (average R2 of 0.60 and average rRMSE of 11%) and high coherence with spatial patterns induced by pasture composition, grazing and topography. Our results confirm the potential of current (e.g., PRISMA, EnMAP) and upcoming operational hyperspectral missions (e.g., ESA's CHIME) for regional forage biomass and nutritional quality assessment.