How to consistently inform NEON’s land surface model with tower-based eddy-covariance flux observations? A novel approach to spatio-temporal rectification
Eddy covariance (EC) observations of ecologically relevant trace gas and energy fluxes are too sparse spatially for direct assimilation into gridded earth system models (ESMs) or for comparison with large-scale observations. The spatial coverage of a tower EC measurement may represent less than 1% of the domains typically covered by ESMs and remote sensing data. It is hence desirable to improve the spatial representativeness and temporal consistency of EC measurements for improving ecological inference. The objectives of this study are (i) to provide consistent flux time-series for target regions, rather than for a spatio-temporally variable patch of surface close to tower sites, and (ii) to test the applicability of the presented procedure across eco-climatological gradients covering some of the complexity of NEON sites.
Based on the environmental response function approach (ERF, Metzger et al., 2013), we developed a procedure to produce spatio-temporally explicit flux fields from tower EC observations. The underlying principle of ERF is to extract relationships between ecological responses and biophysical drivers under varying environmental conditions. For this purpose, spatially gridded land surface observations and high-frequency EC flux processing and footprint realizations (≈60 h−1) are utilized. Provided sufficiently good calibration, the resulting ERF can then be used for projecting the surface-atmosphere exchange to the larger surroundings of the EC measurements.
During our pilot work we apply ERF to EC measurements from July and August 2011 at the AmeriFlux Park Falls tall tower, Wisconsin, U.S.A. For sensible heat flux we find a residual standard error between EC measurements and calibrated ERF as small as 2 W m−2. Moreover, when the ERF is used for projection, the representative spatial coverage can be expanded to >95%, >80% and >70%, for target areas around the tower of 25 km2, 100 km2 and 400 km2, respectively. Based on this, we further validate the applicability of the tower ERF procedure for three additional AmeriFlux sites in distinctly different climate and ecological environments: Niwot Ridge subalpine conifer, Santa Rita mesquite savanna, and Lost Creek shrub fen wetland. Lastly, for each site we determine an uncertainty budget for extending EC tower observations of heat, water vapor, CO2, and CH4 exchange to regional scales. Our results also substantiate the potential of the ERF procedure for improved data assimilation through informing ESMs with flux fields.