Misure di flussi di CO2 con la tecnica eddy-covariance in foresta: analisi di qualità dei dati ed interpretazione della dinamica spazio-temporale dei flussi a scala locale
Author(s)
Virdiano, Giovanni Andrea
Date Issued
March 11, 2011
Type
Doctoral Thesis
Abstract
The eddy-covariance technique (EC) is widely applied for estimating the CO2 fluxes between atmosphere and
terrestrial ecosystems, for monitoring forest ecosystems and its physiology and for climate change-related studies.
This technique is based on the assumption that in a footprint upwind from the measurement point (located over the
canopy), the horizontal mass of air moving over and within the canopy is in a stationarity state with regard to the
upward and down ward turbulent fluxes of a scalar as CO2. In such stationarity condition, the turbulent fluxes of CO2
measured at one point over the canopy is a good estimate of the uniform fluxes in the upwind footprint. A good
turbulence is required to mix the entire air column to avoid accumulation of CO2 under the measurement point
(storage flux1, f_st) and/or lateral transport of CO2.
During night-time, the CO2 turbulent flux should provide an estimate of ecosystem respiration rate (R_N). It is
reported that the CO2 eddy-flux frequently underestimate R_N, expecially for non homogeneous terrain and in stable
(low turbulent) atmospheric condition. Hence, the turbulent flux component, measured over the canopy, needs to be
integrated by the estimation of the storage component, that accumulate below the eddy covariance measuring point.
However, the sum of the turbulent and the storage component of the CO2 flux not always represents the effective
ecosystem respiration. This missing CO2 fluxes is linked to failure of EC assumption due to: a non stationary in CO2
storage or in the above canopy air moving; and/or CO2 escaping along slope through air movement and CO2
concentration horizontal gradients. A typical case is the drainage flow, with cooler air moving down-slope.
To circumvent this possibility of R_N underestimation, conventional data-handling approaches use to remove the data
collected under low-turbulence evaluated using friction velocity (u*) thresholds, where u* is used as a index of
turbulent mixing. The u*-threshold depends on site-specific characteristics.
This study aimed at assessing the spatial and micro-meteorological distribution of R_N values at a forest sites,
deriving the expected R_N using relationships among good-quality R_N values and temperature. The patterns of these
“differences” (R_N measured vs expected, % difference) were evaluated against the spatial context of topography
and vegetation cover to analyse the possible causes of underestimation. This analysis was then compared with the
classical u*-threshold criterion.
In order to compute the expected R_N, we searched for the most-reliable respiration value at night: it was found that
the functional relationship between the maximum of R_N of every night (Rmax-n ) and air temperature (T-air) has a much
better correlation (r =0.64) than the same T-air regression to R_N (u*>0.5 m/s filtered; r =0.38). Nevertheless, R_N
values show large fluctuation, also under middle turbulent condition, expecially for footprint of synoptic winds (over
70% of data).
A factor that could explain the good agreement of Rmax-n with T-air was the hight value of storage flux (f_st) that
compensated the relatively low values of turbulent eddy flux. This means that when Rmax-n was occurring in stable
nights, usually the storage flux was not washed away or displaced by competitive processes (advection, drainage flow,
etc.). Analysis of differences was consistent with that interpretation, but predictions were related to u*-thresholds
depending on footprint and wind direction.
Furthermore, the conventional u*-threshold criterion failed to detect the significant under estimation of R_N, under
prevalent conditions of synoptic winds and often rejected good estimates from Rmax-n, with low u*-class but adequate
f_st compensation.
A single u*-threshold criterion (0.4 m/s), applied to the whole dataset, irrespective of flux footprint, leads to i) a mean
underestimation of R_N by approximately 50 % in the prevalent footprint (transversal to the main slope) that should
be due to a decoupling between above and below canopy turbulence and a preferential path of air drainage; ii) an
“overestimation” of ~30% for the wind directions moving along-slopes, where katabatic wind may originate.
Then, in the studied site, the u*-threshold should be calculated for the specific flux-footprint, with different
preferential air movement and vertical decoupling.
At the studied site, this method of data analysis, based on consistencies and anomalies in the flux database (actual vs.
expected R_N values) finds support also from the spatial analysis of the vegetation (heterogeneity, differences in
structure and plant density, etc) and from expected drainage flow, allowing to discriminate these two causes. In
reaching this goal, the positioning of the EC measurement point with respect to vegetation cover and wind pattern
was essential.
The analysis, here suggested, was implemented by algorithms, integrated in a data treatment procedure. That a
procedure could be useful for recursive data processing or comparative application to other eddy-covariance dataset,
in order to validate and calibrate the procedure for different spatial and dynamic context. Ancillary field
measurements of horizontal CO2 gradients and wind vertical profile could be performed to get more confidence on
predictions of this analysis.
Additional information
Dottorato di ricerca in Ecologia forestale
File(s)![Thumbnail Image]()
Name
gavirdiano_tesid.pdf
Size
25.56 MB
Format
Adobe PDF
Checksum (MD5)
96ad963df52c860717d03eff4dd8dd2d