Dealing with disjunct concentration measurements in eddy covariance applications: a comparison of available approaches.

Using proton transfer reaction mass spectrometry equipped with a quadrupol mass analyser to quantify the biosphere-atmosphere exchange of volatile organic compounds (VOC), concentrations of different VOC are measured sequentially. Depending on how many VOC species are targeted and their respective integration times, each VOC is measured at repeat rates on the order of a few seconds. This represents an order of magnitude longer sample interval compared to the standard eddy covariance (EC) method (5-20 Hz sampling rates). Here we simulate the effect of disjunct sampling on EC flux estimates by decreasing the time resolution of CO and HO concentrations measured at 20 Hz above a temperate mountain grassland in the Austrian Alps. Fluxes for one month are calculated with the standard EC method and compared to fluxes calculated based on the disjunct data (1, 3 and 5 s sampling rates) using the following approaches: i) imputation of missing concentrations based on the nearest neighbouring samples (iDEC), ii) imputation by linear interpolation (iDEC), and iii) virtual disjunct EC (vDEC), i.e. flux calculation based solely on the disjunct concentrations. It is shown that the two imputation methods result in additional low-pass filtering, longer lag times (as determined with the maximum cross-correlation method) and a flux loss of 3-30 % as compared to the standard EC method. A novel procedure, based on a transfer function approach, which specifically corrects for the effect of data treatment, was developed, resulting in improved correspondence (to within 2 %). The vDEC method yields fluxes which approximate the true (20 Hz) fluxes to within 3-7 % and it is this approach we recommend because it involves no additional empirical corrections. The only drawback of the vDEC method is the noisy nature of the cross-correlations, which poses problems with lag determination - practical approaches to overcome this limitation are discussed.