Proximal sensing of forests CO2 uptake

Understanding the relationship between ground-based vegetation spectral signal and eddy covariance-based estimates of gross ecosystem production (GEP) is important for interpretation, evaluation and validation of aircraft and satellite-derived GEP products and thus for further up-scaling of observations. Several vegetation indices (VIs) were developed to monitor different ecosystem’s characteristics strictly related to GEP. Among them, the normalized difference vegetation index (NDVI) and the photochemical reflectance index (PRI) are the most commonly used as indicatives of fraction of photosynthetically active radiation absorbed by vegetation (fAPAR) and light use efficiency (ɛ), respectively. In this study we evaluated the ability of VIs to estimate GEP across three different forest ecosystems (Silver fir-Norway spruce mixed forest in Lavarone, North-East Italy; European larch forest in the Aosta Valley, North-West Italy; and Scots pine forest in Tuczno, North-West Poland) using different Monteith’s light use efficiency model formulations (LUE model assuming that GEP is driven by absorbed photosynthetically active radiation - APAR and ɛ). At each site, continuous tower-based measurements of VIs were performed by Skye multispectral sensors (Skye Instruments, Llandrindod Wells Powys, UK) for a period of at least two years. The CO2 flux data for corresponding years were derived from eddy covariance (EC) measurements. In the larch forest - characterized by strong seasonal dynamics of fAPAR - NDVI proved to be effective in tracking seasonal changes of GEP. On the other hand, in the evergreen forests - where “greenness indices” such as NDVI and thus fAPAR remained relatively constant throughout the whole year - the GEP dynamics showed to be largely controlled by changes in ɛ. In fact, the accuracy of GEP estimations in the analyzed evergreen ecosystems improved when implementing the highly seasonally variable PRI vegetation index in the LUE model formulation