Uncertainties in measurement and modelling of net ecosystem exchange of a forest

At Hyyiala, SMEAR II measurement station in Southern Finland, the eddy covariance (EC) measurements of ecosystem exchange were performed above a Scots pine forest with two closely positioned flux measurement systems. The separation distance of about 30 m between the two systems surrounded by a homogeneous stand of the radius 200 m implies practically the same micrometeorological flux measurement source area, but partial decorrelation and different observation of turbulence and fluxes due to spatial separation. Thus, these fluxes provide two not fully independent estimates of net ecosystem exchange (NEE) of the same forest. NEE, obtained from the EC measurements, was intercompared. with modelled values for periods consisting of several months of observations. The measurements as well as model values were studied for random errors as well as systematic differences. The process-based canopy exchange model was as successful as direct EC measurements in estimation of NEE in terms of random errors and systematic biases in NEE and serves as a useful tool for studying the canopy processes and gap-filling of missing values in measurements. As the EC technique is not free of possible error sources, modelling is also useful for evaluation of flux measurements. The important conclusion of current study is that the main uncertainty of long-term ecosystem NEE does not possibly result from underestimation of night-time observations as commonly believed, but could be instead related to day-time observations. For the given site, the yearly uncertainty of annual NEE was estimated to be 80 gC m(-2) year(-1), originating mainly from the uncertainty of day-time observations. (c) 2006 Elsevier B.V. All rights reserved.