Application

As it is described in the previous section, the NEE calculation requires accurate carbon dioxide storage calculations (Fan et al., 1990) from the concentration profiles. This is generally calculated using a finite difference approach of term 1 in the right hand side of equation :

$$\int ^{z_{r}}_{0}\frac{\partial \overline{c}}{\partial t}dz=... ...c}dz=\frac{\Delta \left( \sum c_{z}\Delta z\right) }{\Delta t}$$ (3.5)

Usually, the profiles are calculated using simple linear interpolation between levels, where the concentration below the lowest level is considered to be constant (K. Davis, pers. comm.).

Similarity theory (see section ) provides a good framework for the NEE calculations in another way. A more accurate storage calculation is performed by fitting the concentration data to obtain the concentration profile using an appropriate similarity function.

In our case the rate of change of the CO storage below the measuring level is calculated as follows:

1. First, the carbon dioxide profile data are interpolated for each hourly boundary datum using a cubic spline function (Horváth and Práger, 1985) (concentration data are available generally in each 8-10 minutes, which is not neccessarily coincides with the beginning of the hours). As a result, two complete profiles are obtained for the beginning and for the end of a specific 60 min period.

2. A theoretical logarithmic CO profile is constructed for the two profiles based on the closest available stability parameter (), friction velocity () and scalar scaling parameter () using eq. .

3. Since the measured CO profile values are considered to be exact, the theoretical logarithmic profiles are modified to fit the measured data. Linear interpolation is used between 82 m and 48 m, since the deviation of the linear profile from the logarithmic profile is slight here. Below 48 m, the logarithmic profile is forced to fit the measured values at 48 m and 10 m. This is accomplished by shifting and stretching the ideal profile to fit the data.

Comparison with the rate of change of storage calculated using the linear method shows small difference, but it is well known that slight systhematic errors may grow to cause considerable bias in the integrated net flux (Moncrieff et al., 1996).

As it was described in section , real mean vertical wind speed is calcualted for our site as part of the wind vector rotation routine. This value can be utilized to calculate the vertical mass flow term for each hourly interval (Lee, 1998) as it is performed by e.g. Baldocchi et al. (2000). However, it was found that this method should not be used for our site. The reason is that during nighttime the 82 m level gets out of the inversion layer and completely decouples from the ground. In such a situation, the onset of a non-zero mean vertical velocity clearly can not advect the whole, 82 m thick air column, but only the upper part of it, which lies above the inversion with considerably lower CO concentration. This results in a clearly different mass flow term compared to the one estimated from the 3rd term in eq. . In fact, inclusion of term 3 in the right hand side of eq. into the calculation of NEE yields ecologically incorrect results: it causes extremely high or low (depends on the day and hour) respiratory rates during nighttime.

If the measuring level lies inside the inversion, the mass flow correction may be more plausible, but the most recent investigations propose that the correction must be treated with caution, as it was described in section (Finnigan, 1999; Yi et al., 2000). During daytime, when the lower boundary layer is well-mixed, term 3 in eq. 0 because of the slight difference between the CO concentration at 82 m and the average CO concentration of the air layer below 82 m.

Summarizing the above, NEE is calculated as the sum of the eddy flux measured at 82 m plus the rate of change of CO storage below 82 m for the large scale system (NEE=). For the Japanese system at 3 m, the eddy flux is considered to be equal to NEE because of the slight storage change below. The interpretation of the flux data calculated by means of the similarity theory is descibed in the next section.