Tower-based optical sensing of ecosystem carbon fluxes
Optical sampling of spectral reflectance and solar induced fluorescence provide information on the physiological status of vegetation that can be used to infer stress responses and estimates of production. Repeated observations are required to observe the effects of changing environmental conditions on vegetation. This study examines the use of optical signals to determine inputs to a light use efficiency (LUE) model describing productivity of a cornfield where carbon flux, spectral reflectance and fluorescence were measured.
Data were collected at the Optimizing Production Inputs for Economic and Environmental Enhancement (OPE3) fields (39.03°N, 76.85°W) at USDA Beltsville Agricultural Research Center. Agricultural Research Service researchers measured CO2 fluxes using eddy covariance. Optical measurements were made from a nearby tower supporting the FUSION sensor system consisting of two dual channel, upward and downward looking, spectrometers simultaneously collecting high spectral resolution measurements of reflected and fluoresced light. The relationships among the different optical measurements indicate that they are providing different types of information on the vegetation and that combinations of these measurements provide improved retrievals of CO2 fluxes.
Seasonal and diurnal changes in corn were observed through both Gross Ecosystem Production (GEP) from eddy covariance and spectral reflectance from FUSION, in this case from a single viewing direction (view zenith angle=25° and view azimuth angle=330° from North). Measurements were connected through the Light Use Efficiency equation:
GEP = e (fAPAR Qin)
where Qin is incident Photosynthetically Active Radiation (PAR), fAPAR is the fraction of absorbed PAR, and e is the light use efficiency. fAPAR is estimated using a linear relationship with NDVI. e is estimated using the Photosynthetic Reflectance Index (PRI), that detects changes in Xanthophyll cycle pigments using reflectance at 531 nm compared to a reference band at 570 nm, and through a Partial Least Squares Regression (PLSR) of spectral bands (400-900nm) sampled at 5 nm using 250 randomly selected points for training the algorithm.
The estimation of half-hourly GEP through the 2013 growing season using only the optical measurements finds that using NDVI for fAPAR combined with incident PAR alone has an R2 of 0.47. Adding in a description of e variability using PRI improves the relationship to a R2 of 0.67. Calculating e using the partial least squares regression improves the R2 to 0.79.