A true eddy accumulation system for trace gas fluxes using disjunct eddy sampling method

. A new true eddy accumulation flux measurement system, based on the disjunct sampling approach, has been developed and tested. In disjunct sampling, short, separate samples are taken instead of continuously sampling the air as in traditional relaxed eddy accumulation and eddy covariance systems. This method reduces the number of samples but allows more time to process them. Simulation shows that the fluxes, calculated using disjunct data, are close to those calculated using continuous data. The disjunct true eddy accumulation instrument was successfully deployed to measure monoterpene fluxes at Niwot Ridge, Colorado. The ability of the system to measure the sample flow accurately, critical for any eddy accumulation system, was tested using atmospheric CFC-113 as a tracer, with good results. The system was capable of measuring relatively low mpinene fluxes, below 10 ng m -2 $-1 at T = 8 ø - 18øC, over a subalpine forest.


Introduction
The For measuring the vertical fluxes of those trace gases for which no fast sensor exists, two approaches are used. The first and traditional approach is the gradient method, in which the flux is inferred from the mean concentration difference between two or more measurement heights. Variants of this method have been widely used for measurements of biogenic hydrocarbon fluxes [e.g., Fuentes et al., 1996;Guenther et al., 1996 Traditional flux measurement methods sample air in a continuous manner. It is, however, also possible to sample noncontinuously and still have enough samples to calculate the fluxes [Haugen, 1978;Kaimal and Gaynor, 1983;Lenschow et al., 1994]. In this approach, called disjunct sampling by Lenschow et al. [1994], quick samples are taken with a relatively long timeinterval between them (Figure 1). This gives more time to process the samples, but increases the statistical uncertainty of the fluxes calculated using such disjunct data. As a large part of the fluxes in the atmospheric surface layer is carried by relatively large eddies, continuously sampled samples are not totally independent. Thus the statistical uncertainty of the fluxes does not increase as much as they would in the case of independent samples. According to Lenschow et al. [1994], the disjunct sampling does not increase the statistical uncertainty of the flux values more than 8% if the time interval between samples is shorter than the appropriate integral timescale.
In this paper we present a system for disjunct true eddy accumulation (

Sampling
In the true eddy accumulation method [Desjardins, 1977], air is sampled into updraft and downdraft reservoirs proportionally to the vertical wind velocity. The A disjunct eddy accumulation system quickly grabs samples (sampling time, ts < 0.5s) every 10-60 s into an intermediate storage reservoir (ISR, Figure 3). When the sample is in the ISR, a small part of it is collected into the updraft or downdraft sample container, according to the sign of the vertical wind velocity at the moment the grab sample was taken, and proportionally to the vertical wind velocity. After the sample has been collected, the ISR is evacuated for the next round. The sampling sequence is repeated typically for 30-60 min, long enough to sample eddies of all scales. The equation for true eddy accumulation with disjunct sampling can be written re = * + (w-)d.
The notation of averages in (4) and (6) is the same as in (1)and (2).

Description of the Instrument
The instrument described in this paper consists of two ISRs (Figure 4), which are operated by turns. When one is being evacuated, the other one is sampling. This way it is possible to get twice as many samples as with one ISR. The operation sequence of the system is de- According to the simulation the underestimation caused by the carryover was around 4% and the correlation coefficient r 2 between fluxes with and without carryover

Experiment
The field test of the DEA system was conducted at the Ameriflux site near Niwot Ridge, Colorado (40ø02'N, 105ø33'W, 3050 m above sea level) in August-October

The forest surrounding the measurement tower consists mainly of Engelmann spruce (Picea engelmannii Parry ex Engelm.), lodgepole pine (Pinus contorta Dougl. ex Loud.), and subalpine fir (Abies lasiocarpa (Hook.). Nutt.). The zero plane displacement height
at the site is 12 m. The site has 5 ø slope eastward which was taken into account by rotating the wind vector. The rotation was done using predetermined rotation angles in such a way that the w axis was perpendicular to the mean streamlines. The rotation angles were determined from previous measurements made at the site. The vertical wind data were not high-passfiltered. High-pass filtering could remove the effect of tilted mean streamlines on the mean vertical wind, and longer-term horizontal gusts on the vertical wind variation, but the effect of short-term variation would remain. The measureznents were conducted on a 30 m high walk-up tower. The height of the DEA was 19.3 m, and that of the acoustic anemometer was 20.3 m.  Supelco, Bellefonte, Pennsylvania). The cartridges were analyzed using a gas chromatograph with mass spectrometer. A more detailed discussion on the chemical analysis is given by Greenberg et al. [1999]. The cartridges were stored at -30øC until analyzed.
The temperatures at the site during measurements were relatively low, ranging from 2øC to 18øC. The typical winds were westerly, but in the afternoons some easterly upslope winds occurred.

Results and Discussion
A critical aspect for a cartridge eddy accumulation system is the ability to measure the sample volume accurately. To test how well the sample volumes were measured, the ratio of CFC The data provided by DEA method, as well as any micrometeorological methods, need to be filtered to remove the data for which the micrometeorological and other assumptions do not hold. This filtering should be done with objective criteria by setting limiting values for suitable parameters. Generally, friction velocity has been used to exclude the data with too weak turbulence by setting a lower limit of u,=0.1 m s -1. The use of normalized mean vertical wind Wn to exclude the data with too tilted mean wind is discussed above. The importance of integral timescale is presented by Lenschow et al. [1994]. Also, stationarity seems to be an important factor, but an objective filtering parameter needs to be developed.
A major advantage of disjunct true eddy accumulation lies in the fact that the true eddy accumulation is a direct flux measurement method and does not de-pend on similarity assumptions or empirical relationships. This makes the approach less wfinerable to systematic errors, which can affect gradient and REA measurements.
Future applications of the disjunct eddy sampling approach can include several different flux measurement techniques. The disjunct sampling strategy can enable eddy covariance measurements of trace gas fluxes using semifast sensors with response times between 1-60 s. Also, different variants of eddy accumulation method, such as hyperbolic relaxed eddy accumulation [Bowling et al., 1999], can be incorporated with the disjunct sampling approach. These applications could be used for flux measurements of a wide range of che•nical compounds. The disjunct eddy sampling method also has potential for aircraft applications since the wind data, which can be delayed for a few seconds, are easily accommodated by the technique.

Concluding Remarks
The disjunct sampling approach can be incorporated into several eddy flux measurement techniques, true eddy accumulation described in this paper being only one of them. The instrument described in this paper was able to measure low fluxes of c•-pinene at Niwot Ridge, Colorado. Data filtering using objective exclusion limits was used to ensure data quality. Fluxes measured by the disjunct true eddy accumulation instrument were in the lower range of expected emissions, below 10 ng rrt -9• s -• at T = 8 ø -18øC.
The simulations of disjunct sampling by continuous w and T time series confirmed conclusions of Lenschow et al. [1994], that disjunct sampling does not greatly increase the statistical uncertainty of flux values. The critical sample volume measurement was tested using atmospheric CFC-113 as a tracer. Results showed very close agreement between updraft to downdraft ratios of CFC-113 peak area and sample volume. The major advantage of the DEA lies in the fact that it is a direct flux measurement method. Further efforts are needed to characterize the accuracy and reliability of this method relative to other flux measurement techniques.