Abstract/Summary

Eddy-covariance-based estimates of net ecosystem exchange are subject to various sources of systematic bias and random measurement uncertainty. Here we concentrate on the cumulative effect of random uncertainty on annual estimates of net ecosystem productivity of carbon (NEP). A 8-year data set of eddy covariance measurements over a mixed deciduous forest at the Morgan-Monroe State Forest (MMSF, Indiana, USA) was used, in conjunction with a 6-day period of paired observations with the AmeriFlux portable system, to evaluate two different approaches to estimate measurement system uncertainty, and an analogous method to estimate the uncertainty in a standard parametric model used to fill data gaps in the annual time series. The cumulative annual uncertainty was obtained by Monte Carlo simulation, separately for the observations and the model estimates. Our results indicate that the overall uncertainty of annual NEP is dominated by the contribution of the gap-filling model, even at relatively small gap fractions of 20%. The magnitude of random uncertainty in NEP varied between ±10–12 gC m−2 yr−1 (i.e., 3–4% of annual NEP at MMSF for years 1999–2006). Thus it must be expected that random uncertainty of eddy-covariance-based NEP is small compared to other potential sources of systematic bias, but we note that very little is known about the long-term cumulative covariate effects of systematic bias in the measured flux.