Estimating the uncertainty in annual net ecosystem carbon exchange: spatial variation in turbulent fluxes and sampling errors in eddy‐covariance measurements

Abstract

Above forest canopies, eddy covariance (EC) measurements of mass (CO 2 , H 2 O vapor) and energy exchange, assumed to represent ecosystem fluxes, are commonly made at one point in the roughness sublayer (RSL). A spatial variability experiment, in which EC measurements were made from six towers within the RSL in a uniform pine plantation, quantified large and dynamic spatial variation in fluxes. The spatial coefficient of variation (CV) of the scalar fluxes decreased with increasing integration time, stabilizing at a minimum that was independent of further lengthening the averaging period (hereafter a 'stable minimum'). For all three fluxes, the stable minimum (CV 5 9-11%) was reached at averaging times (s p ) of 6-7 h during daytime, but higher stable minima (CV 5 46-158%) were reached at longer s p (412 h) during nighttime. To the extent that decreasing CV of EC fluxes reflects reduction in micrometeorological sampling errors, half of the observed variability at s p 5 30 min is attributed to sampling errors. The remaining half (indicated by the stable minimum CV) is attributed to underlying variability in ecosystem structural properties, as determined by leaf area index, and perhaps associated ecosystem activity attributes. We further assessed the spatial variability estimates in the context of uncertainty in annual net ecosystem exchange (NEE). First, we adjusted annual NEE values obtained at our long-term observation tower to account for the difference between this tower and the mean of all towers from this experiment; this increased NEE by up to 55 g C m À2 yr À1 . Second, we combined uncertainty from gap filling and instrument error with uncertainty because of spatial variability, producing an estimate of variability in annual NEE ranging from 79 to 127 g C m À2 yr À1 . This analysis demonstrated that even in such a uniform pine plantation, in some years spatial variability can contribute $ 50% of the uncertainty in annual NEE estimates.

Uncertainty analysis of eddy covariance CO<sub>2</sub> flux measurements for different EC tower distances using an extended two-tower approach

“…When applying a two-tower based approach it is important to assure that systematic differences of the measured fluxes, which are partly caused by within-site or among-site heterogeneity, are not attributed to the random error estimate of the measured NEE. Our assumption that even within a site with apparently one uniformly distributed vegetation type (and for very short EC tower distances) land surface heterogeneity can cause significant spatial and temporal variability in measured NEE is, e.g., supported by Oren et al (2006). They found that the spatial variability of ecosystem activity (plants and decomposers) and leaf area index within a uniform pine plantation contributes to about half of the uncertainty in annual eddy covariance NEE measurements while the other half is attributed to micrometeorological and statistical sampling errors.…”

“…It is expected that systematic differences in measured NEE caused by spatially variable land surface properties are stronger during the night than during the day since they affect respiration more directly than photosynthesis (see, e.g., Oren et al, 2006). Moreover, during night-time and/or winter (positive NEE), some conditions associated with lower EC data quality such as low turbulence, strong stability, and liquid water in the gas analyzer path prevail more often than in summer and/or daytime (negative NEE).…”

“…For larger window sizes (24 or 48 h) further analysis was hampered by too many data gaps -i.e., the 50 % criterion was hardly ever fulfilled and not enough averages remained to allow for the two-tower based uncertainty estimation (Table A2). The correction was done separately for positive and negative fluxes, due to the different sources, properties and magnitudes of the CO 2 flux measurements and different errors for daytime (negative) and night-time (positive) fluxes (e.g., Goulden et al, 1996;Oren et al, 2006;Wilson et al, 2002). The final sfd-corrected NEE1 corr values for EC1 and NEE2 corr values for EC2 should not be understood as corrected NEE flux data.…”

Energy, water, and carbon fluxes in a loblolly pine stand: Results from uniform and gappy canopy models with comparisons to eddy flux data

“…The stem density is derived based on the assumption that the crowns are randomly distributed in space, and the LAI is derived based on the assumption that the foliage volume density is 0.7 m 2 /m 3 . ments, (2) random variation from flux tower measurements due to the dynamic nature of the flux footprint [Oren et al, 2006], and (3) errors in model parameters in PenmanMonteith equation. If we consider the daytime trend for the entire growing season, LE from GCR better matches with the measured LE than that from UCR ( Figure 7).…”

“…If we consider the daytime trend for the entire growing season, LE from GCR better matches with the measured LE than that from UCR ( Figure 7). When comparing the ensemble mean diurnal trends, we essentially removed the high frequency random errors from the eddy flux measurements, which could account for 50% of the variation at the half hourly time step [Oren et al, 2006].…”