Spatial Representativeness and Uncertainty of Eddy Covariance Carbon Flux Measurement for Upscaling Net Ecosystem Productivity to Field Scale

Abstract:

Eddy covariance (EC) technique is considered one of the most direct, defensible ways to measure and calculate turbulent fluxes within the atmospheric boundary layer and is often assimilated into biogeochemical model to constraint the model parameters or as a reference is used to validate the estimated net ecosystem productivity (NEP) from satellite remote sensing and biogeochemical model. However, EC measurement representing an integrated flux over its footprint area which is non-match with model grid or remote sensing pixel. Quantifying the uncertainties associated with gridded flux estimates by upscaling single EC tower NEP measurement to pixel scale is an important but not full investigated issue due to data availability. Heihe Watershed Allied Telemetry Experimental Research (HiWATER) Multi-Scale Observation Experiment on Evapotranspiration (MUSOEXE) built a flux observation matrix that include 17 EC towers within 5 km × 5 km area, provides an unprecedented opportunity to evaluate this uncertainty.

This study evaluated the spatial representativeness and uncertainty of EC measurement for upscaling to field scale combine footprint model, VPRM model and remote sensing data. Results shows that the large spatial variability of GPP, Re, and NEP is existing within test field (cropland landscape) during the growing season from 10 June to 14 September 2012. These variability increase with the increase of GPP and NEP. The systematic underestimations of single EC tower may exceed 100%, 40%, and 300% and the overestimations may exceed 35%, 30%, and 70% in extreme cases for GPP, Re, and NEP, respectively. This illustrate the risk of single tower EC measurement be used to validate remote sensing NEP product at global scale by direct comparison. This systematic bias is dominated by the different of vegetation structural component between two scales for NEP and lead to a systematic bias to estimate field mean NEP by impact the VPRM parameters. A simple linear model was proposed to reduce the representation error of single EC tower that combined the vegetation structure information. The corrected EC based NEP measurement then be used to constraint VPRM parameters to estimate the field mean NEP. The availability is validated by compare with reference value estimated using a linear unmixing model proposed in this study.