Hyperspectral Remote Sensing for Seasonal Estimation of Aboveground Biomass in Grassland Habitats

Dry grassland sites are amongst the most species rich habitats of Central Europe. In Switzerland, they are home to a large number of plant and animal species that are classified as endangered or threatened. A key component for designing optimal and effective management schemes ensuring the sustainability of these ecosystems, is knowledge of their biomass production. In this study we explored the potential of hyperspectral remote sensing for mapping above-ground biomass in grassland habitats along a dry-mesic gradient, independent of a specific habitat or a phenological period. We developed statistical models between spectral reflectance collected with a spectrometer but resampled to Hyperion bands, and biomass samples. We then tested to what degree the calibrated biomass models could be scaled to actual Hyperion data collected over the study area. Biomass samples (n = 155) were collected from 11 grassland fields located in the Central part of the Swiss Plateu. To capture normally occurring variation due to canopy growth stage and management factors sampling was performed at 4 time steps during the 2005 growing season. We investigated the relationship between biomass and a.) existing broad and narrow-band vegetation indices, b) narrow band NDVI type indices and c.) multiple linear regression using branch-and-bound variable search algorithms. Best models for estimating and predicting biomass were obtained from the multiple regression and narrow band NDVI type indices contrary to existing vegetation indices. Spectral regions related to plant water content were identified as the best estimators of biomass. Furthermore, results from this study demonstrated the importance of seasonal biomass measurements for building reliable models. Finally, promising results in estimating grassland biomass were not only obtained for the Hyperion resampled field spectrometer data, but also for the actual Hyperion data, showing the potential of up-scaling to the landscape level.