Objective 1
We aimed to verify which vegetation and environmental factors are the most important in determining the spatial and temporal variability of GPP and LUE. The impact of different potential drivers (i.e. temperature and water-related variables, plant traits) of GPP and LUE were analyzed by using statistical methods. The main results of this analysis has been published by the researcher in Global Ecology and Biogeography (Balzarolo et al., 2019). This analysis revealed that drought-related variables had an impact on annual light use efficiency in temperate forests, while temperature-related variables affected annual of cold forest. Furthermore, the seasonal variability of LUE was strongly related to the meteorological variables and varied along the season and was stronger in summer than spring or autumn.
Objective 2
Satellite data available from different platforms (i.e. MODIS, Proba-V and Sentinel-2) were collected a different remote sensing indicators were calculated. We focused mainly on remotely-based indicators that reflect canopy functioning and physiology (i.e. chlorophyll content index−CCI, Photochemical Reflectance Index−PRI, green NDVI, Red-edge) rather than conventional indicators of canopy greenness (i.e. NDVI and fAPAR). The values of each RS indicator was regressed against GPP values, obtained by the eddy covariance technique, for the same days of image acquisitions in order to define the best fitting-functions in different environmental conditions (i.e. drought). We found limitation of MODIS and in situ NDVI in the capture of GPP seasonal dynamics for evergreen forests due to insufficient description of plant ecophysiology by the NDVIs (Balzarolo et al., 2019, published in Remote Sensing). MODIS NDVI and CCI identified the spatial variability of average annual GPP (Fernández-Martínez et al, 2019, published Remote Sensing).
Objective 3
The performance in predicting annual and seasonal GPP variability has been validated against in situ observations available from global databases (i.e. FLUXNET) and benchmarked against other remotely-based global models (e.g. MOD17). Furthermore, to improve the GPP upscaling and validation by in situ observations we focused on the analysis of the impact of landscape spatial heterogeneity on the relationship between GPP and RS indicators. The best performances were obtained for the highest spatial resolution (i.e. at 0.5 × 0.5 km for MODIS) (Balzarolo et al., 2019; Remote Sensing). Furthermore, we demonstrated that the landscape heterogeneity estimated using the spatial heterogeneity index is essential in upscaling in situ GPP data.
The results of the project were presented at scientific conferences and symposia and at meetings with potential stakeholders, as well on peer-reviewed scientific journals (see the section Publications).