Estimate of turbulent fluxes with eddy-covariance technique in a complex topography: A case study in the Italian Alps

A sensitivity analysis to different eddy—covariance data processing algorithms is presented for a dataset collected in an Alpine environment with complex topography. In Summer 2012 a micrometeorological station was installed at Cividate Camuno (274 m a.s.l., Oglio river basin, Central Italian Alps), in a flat and rectangular grass-covered lawn. The grass was 0.6 m tall during most of the field campaign. The station is equipped with traditional devices, four multiplexed TDR probes, and an eddy--covariance apparatus sampling at 20 Hz (Gill WindMaster Sonic Anemometer and Licor Li7500 Gas Analyzer), at about 3 m above the ground. The local winds regime is strongly affected by the morphology of the valley, and the topography is complex also due to the heterogeneity of the surrounding-areas land—cover. Using EddyPro software, the sensitivity of the turbulent fluxes estimate was assessed addressing three major issues of the data processing procedure, i.e. the choice of the computational averaging period, of the axis rotation method and of the data detrending criterion. Once identified three test periods of consecutive days without rainfall, the fluxes of momentum, sensible heat and latent heat were computed at the averaging period of 30, 60 and 120 min respectively. At each averaging period, both the triple rotation method, the double rotation method and the planar fit method were applied. Particularly the latter was applied both fitting a unique plane for all the wind directions and fitting multiple planes, one for each sector of the wind rose. Regarding the detrending criteria, data were processed with a block average and a linear detrend, the latter with time constant of 5, 30, 60 and 120 min respectively. Therefore, for each test period about 50 estimates of the fluxes were provided. As a result the obtained fluxes were compared. Even if with different flux quality, their pattern is quite stable with regard to the applied estimate procedures, but with sensitively different average values.