Accurate CO2 concentration gradient measurements are needed for the computation of advective flux terms, which are part of the full Net Ecosystem Exchange (NEE) budget equation. A typical draw back of current gradient measurement designs in advection research is the inadequate sampling of complex flow phenomena using too few observation points in space and time. To overcome this draw back, a new measurement design is presented which allows the parallel measurement of several sampling points at a high frequency. Due to the multi-analyzer nature of the design, inter-instrument bias becomes more of a concern compared to conventional setups. Therefore a statistical approach is presented which allows for accurate observations of concentration gradients, which are typically small in relation to analyzer accuracy, to be obtained. This bias correction approach applies a conditional, time dependent signal correction. The correction depends on a mixing index based on cross correlation analysis, which characterizes the degree of mixing of the atmosphere between individual sample points. The approach assumes statistical properties of probability density functions (pdf) of concentration differences between a sample point and the field average which are common to the pdf’s from several sample points. The applicability of the assumptions made was tested by Large Eddy Simulation (LES) using the model PALM and could be verified for a test case of well mixed conditions. The study presents concentration time series before and after correction, measured at a 2m height in the sub-canopy at the FLUXNET spruce forest site Waldstein-Weidenbrunnen (DE-Bay), analyzes the dependence of statistical parameters of pdf’s from atmospheric parameters such as stratification, quantifies the errors and evaluates the performance of the bias correction approach. The improvements that are achieved by applying the bias correction approach are one order of magnitude larger than possible errors associated with it, which is a strong incentive to use the correction approach. In conclusion, the presented bias correction approach is well suited for – but not limited to – horizontal gradient measurements in a multi-analyzer setup, which would not have been reliable without this approach. Finally, possible future improvements of the bias correction approach are outlined and further fields of application indicated.
powered by BayCMS © 2015 University of Bayreuth,