Carbon dioxide exchange of an alpine grassland: integration of eddy covariance, proximal sensing and models

The terrestrial biosphere represents a large pool of carbon, whose cycle is governed by the opposed processes of CO2 uptake (photosynthesis) and release (respiration) from and to the atmosphere. Considering the role of carbon dioxide in the observed global warming, monitoring, understanding and modeling carbon exchange of ecosystems is a critical issue in climate change researches. Moreover because of the multiple implications of vegetation structure dynamics on ecosystem carbon fluxes, monitoring and modeling plant phenology is also of increasing scientific interest. Among terrestrial ecosystem grasslands cover almost 40% of ice-free land surface, nevertheless their role as sources/sinks of atmospheric CO2 is not well clarified. In this study the eddy covariance method was used to assess CO2 exchange at an high elevation unmanaged grassland in the North-Western Italian Alps (Aosta Valley - Torgnon), during three years (2008-2010) of measurements and to evaluate how environmental factors affect photosynthetic processes. The seasonal and inter-annual course of net ecosystem CO2 exchange (NEE), ecosystem respiration (Reco), gross primary production (GPP) and the main meteorological variables was analysed. The three growing seasons had a similar seasonal dynamic, characterised by a fast rise of photosynthetic activity after snow-melt followed by a gradual autumnal decline. Regarding the meteorological variables, only precipitation, soil water content and snow depth differed markedly among two of the studied years (2009-2010) compared to other factors which showed only small differences in restricted time-periods. To better interpret how weather variables modulate ecosystem processes at multiple time-scales (day, week, month, year), a quantitative analysis was performed applying wavelet coherence between time-series of GPP and time-series of different meteorological factors (air and soil temperature, soil water content and photosynthetically active radiation). Eddy covariance and meteorological data were combined with proximal sensing measurements to identify links between optical indices, canopy development and fluxes. In particular a colour index derived from continuous digital imagery (i.e. Greenness Index, (GI), based on RGB channels) and indices derived from an HyperSpectral System (Hyperspectral Irradiometer, HSI) were used as input to simulate GPP, based on a light use efficiency (LUE) model. Results showed that a LUE model driven by optical indices and meteorological variables is able to describe the GPP trend in the two years of study. In particular the use of different model formulations provided insights on the role of the main meteorological factors controlling grassland photosynthesis. The comprehension of these relationships at stand level is essential for extrapolating such information at different spatial scales.