Bridging the gap between historical and ICOS ecosystem flux data series: methodological choices and their impact on net fluxes

Long-term flux data series are at the heart of ICOS research infrastructure, and multiple ecosystem ICOS stations are highly valued for their historical dataset, which extends years before the ICOS collection. However, instrumental and methodological innovation has led to potential differences in computed turbulent net fluxes, which can result in discontinuities in the datasets. We worked on quantifying the differences between the custom PI-processed historical data and the standardised ICOS-processed data, caused by changes in methodology only. We focused on the spectral correction, as it appeared to be the main cause behind the divergences. The resulting analysis highlighted that the use of certain default options in the ICOS processing pipeline might not be suitable for all site typologies. An example is the denoising of spectra, which is shown to potentially remove true signal, thus inducing artificially high correction factors especially for sites measuring close to the canopy, as their spectral content shifts towards the high frequencies. Once different processing options have been aligned, a more fundamental question, regarding the choice of using spectra rather than co-spectra when performing the correction procedure, is investigated. The effects of both choices are assessed. These findings open the way to possible refinements in the standard ICOS flux processing methodology, and potentially in the widely used flux calculation software.