SPECTRAL ANALYSES OF LONG-TERM MEASUREMENTS OF TURBULENT EXCHANGE OVER MIXED HARDWOOD FORESTS

摘要

It is well recognized that losses in measured eddy-covariance fluxes can be caused by path/line averaging, sensor separation, inadequate sensor frequency response, damping through tubes in closed-path gas analyzers, data processing, etc., (Moore, 1986). Previous studies using spectral methods have shown that corrections to measured fluxes range from a few to over 30% (Eugster & Senn 1995, Leuning & Judd 1996, Horst 1997, Massman 2000). The magnitudes of such flux losses in the higher frequency range could be on the same order as unaccounted lower frequency fluxes due to inadequate averaging time (Rissmann & Tetzlaff 1994, Sakai et al. 2001, Finnigan et al. 2002). Spectral methods used to correct the flux losses often require knowledge of the true spectra and cospectra, as well as the transfer functions (Moore, 1986; Massman, 2000). The most used spectral and cospectral models are those of Kaimal et al. (1972) and Wyngaard and Cote (1972) derived from observations over smooth surfaces and flat terrain. Sakai et al. (2001) indicated that the normalized cospectra of momentum and sensible heat in the roughness sublayer over forest canopies have nearly identical form in convective conditions. Here we examine spectra and cospectra charac-teristics to determine appropriate similarity forms of spectral and cospectral for subsequent applications. Then, we focus on the frequency loss of COz and water vapor fluxes due to damping through the long tubes, assuming this to be the major cause of frequency losses. Variations in the coefficients of damping function and the magnitudes of flux correction are also discussed. Data used here were collected at the UMBS (University of Michigan Biological Station in lower northern Michigan) AmeriFlux site. Eddy-covariance systems (CSAT-3 sonic anemometers and a LiCor-6262 closed path infrared gas analyzers) were installed at 34 m and 46 m. The mean tree height is 22 m and peak LAI is 3.5. Results are derived from over 2,400 hourly spectra and cospectra during June-August of 1999 and 2000 at 46 m. Similar analyses are to be performed for the MMSF (Morgan-Monroe State Forest in south central Indiana) site.