Free-air CO2 enrichment (FACE) of a poplar plantation: the POPFACE fumigation system


Author for correspondence: Franco Miglietta Tel: +39055301422 Fax: +39055308910


  •  A new design of free-air CO2 enrichment (FACE) is presented that has been used to expose a poplar plantation to elevated atmospheric CO2 concentrations in other-wise unaltered conditions, in the open.
  •  This system releases pure CO2 at high velocity, through a large number of small gas jets, causing rapid mixing between CO2 and air. The theoretical and practical aspects of this design are described, with emphasis on the fluid mechanics of air–CO2 mixing in sonic jets. Field performance data, including spectral analysis of short-term fluctuations in CO2 concentrations as well as temporal and spatial CO2 control, are reported for the European project POPFACE facility.
  •  Temporal and spatial performances of the operational POPFACE systems were adequate with average long-term CO2 mole fractions on target. Averages over 1 min measured in the centre of the rings were within ±20% and ±10% of the target concentration for > 91% and > 75% of the time, respectively.
  •  The data presented provide convincing evidence that a pure-CO2 FACE system can achieve reliable control, in terms of the quality of the CO2 control, with significant simplification of construction and reduced capital cost.


The POPFACE project was recently funded by the Commission of the European Union (Directorate General Research) in the frame of Terrestrial Ecosystem Research Initiative (Scarascia-Mugnozza et al., 2000). The main objective of this collaborative project is to investigate the response of a poplar plantation and a number of poplar species to the expected increase in atmospheric CO2 concentrations. For this purpose, the POPFACE project implemented a large-scale Free Air CO2 Enrichment (FACE) facility in Tuscania (Viterbo), in Central Italy.

FACE is a technique for exposing crops, forest plantations and natural vegetation to elevated atmospheric CO2 concentrations without an enclosing structure. FACE technology has developed considerably since the first experiences made by Harper and coworkers in the 1970s (Harper et al., 1973) and by van Mooi and coworkers in the 1980s (Mooi, 1985). At present more than 20 FACE sites are operational around the world in Northern and Central America, in Europe, Asia and Oceania. The size of the FACE plots varies from one meter diameter of the MiniFACE (Miglietta et al., 1996; Miglietta et al., 2001) to the 30 m of the larger FACE systems that have been used to fumigate with CO2 patches of forest plantations (Hendrey, 1999).

FACE experiments are almost unanimously considered to provide the best opportunity to expose patches of managed or unmanaged vegetation to conditions of elevated atmospheric concentrations with minimal alteration of the natural environment where plants are growing. Nevertheless, FACE systems also suffer from some experimental limitations such as the presence of substantial infrastructure, the unavoidable presence of CO2 concentration gradients along the wind direction and short-term fluctuations in CO2 concentration. The use of blowers requires large pipes to allow the circulation of large volumes of air, higher power requirement and significant infrastructure. Moreover, the use of the blowers implies the construction of control rings where everything is operated in the same way as in the FACE, with the only exception of the injection of carbon dioxide.

Recently different groups in the USA (S. Roberts, pers. comm.), and Japan (Okada et al., 2001) have attempted to modify the design of their FACE systems to introduce the release of pure carbon dioxide instead of an air-CO2 mixture. Such ‘pure-CO2’ FACE was thought to be a possible alternative solution to the more conventional systems, providing some potential advantages. This is also the idea that inspired the design of the FACE facility that was used for the European project POPFACE and that will be discussed in this paper.

In the POPFACE system, pure carbon dioxide is released to the atmosphere through a very large number of small gas jets, at high velocity. Such innovation was introduced because the theory of fluid mechanics says that when a gas jet reaches sonic velocity, a shock-wave is created at the jet outlet and the air-CO2 mixing is greatly enhanced (Munson et al., 1990). This paper describes both the theory and the practical aspects of this type of FACE design with emphasis on the fluid mechanics of air-CO2 mixing in sonic jets. But it also reports the data of the field performance of a replicated POPFACE system over the first experimental season. Performance data include the spectral analysis of short-term fluctuations in CO2 concentrations in the FACE plot as well as temporal and spatial CO2 control.

Materials and Methods

The relationship between pressure and CO2 flow rate from gas jets

The first question to be answered while designing a pure-CO2 FACE system was the number, size and spatial arrangement of the CO2-venting jets. The basic idea was to utilize a horizontal pipe holding a given number of tiny holes and change the pressure inside the pipe to regulate the CO2 flow rate. During the initial phase of prototype development changes in CO2 flow rate from a number of jets of different diameter were investigated by measuring the relation between mass flow (g CO2 min−1) and the absolute pressure inside the pipe (MPa). During the measurements, the pressure was changed in 8 steps from 0.15 MPa to 0.45 MPa and the flow rate was calculated gravimetrically by monitoring weight changes of a CO2 cylinder.

Measurements of air-CO2 mixing

Air-CO2 mixing in the vicinity of gas jets at sonic velocity was then investigated by means of both direct measurements and modelling. The measurements were made in the laboratory using a polyethylene pipe of 20 mm internal diameter and 1 mm wall thickness, an Infra-red Gas Analyser with a 0–100% CO2 concentration range (IRGA 1: ADC 0–100%, UK), a second IRGA with 0–2000 µmol mol−1 range (IRGA 2: Li-Cor, Mod. 6252, Lincoln, Nebraska, USA) an air piston pump with a flow rate of 19 l min−1 (Mefar, mod. 8523, I) and an axial blower with a capacity of 10 m3 min−1. A hole of 0.3 mm diameter was drilled onto the pipe and this was arranged horizontally at 100 cm above ground. The CO2 was released through the hole by increasing the pressure in the pipe by means of a manually controlled pressure regulator. The blower was used to create a constant turbulent flow of air perpendicular to the pipe with air movement at about 1ms−1 measured with an hot wire anemometer (TSI Velocity Cal, Mod. 8355, City, MN, USA). The measurements were taken with both IRGAs on a vertical plane normal to the jet, using a very thin probe of 1 mm diameter that was moved at different positions in the proximity of the venting hole. Data were taken by pumping through tubing alternately from IRGA 1 and IRGA 2, depending on the mean concentration observed. The high flow rate of the pump lead to a negligible sampling delay. Absolute concentrations at each individual point on the vertical plane were calculated as one-minute average of data taken every second. A sketch diagram of the measurements made is shown in Fig. 1.

Figure 1.

Details of the methodology used to monitor CO2 concentrations in the proximity of a releasing gas at sonic velocity.

Simulation of air-CO2 mixing

A finite-element Computational Fluid Dynamics model (CFD) was also developed to understand better the mechanisms of air-CO2 mixing in high velocity jets. Computational fluid dynamics modelling requires the definition of a large number of input variables and parameters dealing with turbulence kinetic energy and dissipation, flow and the transport of scalars. Detailed description of model parameters and assumptions will go beyond the scope of this paper and, accordingly, only a brief introductory description of the model will be given here. The CFD model required the definition of a three dimensional domain that was designed using appropriate Computer Aided Design software (Geomesh vs 3.5, Fluent Inc., (Lebanon, NH, USA)) and meshed using a tetrahedral grid. The meshed volume provided a 3D model of the horizontal pipe and of the convergent nozzle. Pressure in the pipe and the airflow in the three dimensional space were variable input boundary conditions that could be adjusted before every simulation run. Navier-Stokes equations were solved for pressure, velocity and flow using the k-e turbulence model. When converged to an optimal solution, the model provided detailed information on a number of variables in the 3D domain, including velocity fields and vectors, CO2 concentrations, static and dynamic pressures, turbulence and energy. All these variables could be easily displayed in computer graph, for visualization and exported in the form of grid referenced data. The CFD model of the jet was validated using the experimental data obtained in the laboratory. The model was implemented using Fluent 4.0 (Fluent Inc.).

Construction details of the POPFACE installation

The POPFACE system was initially developed as a prototype and finally assembled in its operational configuration. It was based on directionally controlled release of pure CO2 from jets located on a number of horizontal pipes forming an octagon (Fig. 2). The operational system was built in three replicates and each FACE octagon enclosed an area of c. 350 m2 with a diagonal of 22.2 m. The size of the octagon was designed to provide an internal area unaffected by the geometry of the CO2 release and with a circular shape and diameter of 16 m. The definition of the internal area originates from the intersection of the radii shown in Fig. 2. The pipes located on the perimeter of the octagon were suspended at the height of the poplar canopy using eight telescopic poles located at the vertices of the octagon. The poles allowed easy and rapid vertical movements of the pipes to account for the rapid changes in the height of the poplars. At the beginning of the experiments, immediately after planting, the pipes were located at c. 50 cm above ground and were moved to follow rapid canopy development throughout the growing season. The operational principle of the directional control of the CO2 injection was that of releasing the gas from the three sides of the octagon located on the upwind side of the experimental plot. This was obtained using a control system equipped with solenoid on/off valves (DANFOSS, s.r.l., Torino, Italy) that could open and close the flow of CO2 to each individual side of the octagon. The valves were located near the central mast where the IRGA, anemometer and control unit were also located (Fig. 2). At the beginning of the project, each side of the octagon consisted of two paired pipes carrying each a different number of jets (Fig. 3). This allowed the release of different amounts of CO2 from the pipes perpendicular to the wind direction and those positioned at 45° thus avoiding the so called ‘carry-over effect’ that would have been caused by an even CO2 release from every pipe in the upwind sector of the FACE.

Figure 2.

Layout of the POPFACE octagon illustrating the spatial arrangement of the different system components (legend in the Figure). The shaded area at the centre of the octagon indicates the area where uniform CO2 enrichment can be obtained.

Figure 3.

The operational principles used by POPFACE to account for changes in the wind direction. Each side of the octagon is made by two-layers of pipes. Pipes carrying 70% more jets are indicated. Opened and closed sides are indicated in the figure for two different situations that correspond to two different wind directions (indicated by the solid arrow).

The horizontal 20 mm internal diameter polyethylene pipes with a length of 8.5 m were manually drilled with 350 and 500 holes of a mean diameter of 0.3 mm. In the subsequent season (year 2000), the jets were drilled using laser technology. This very accuarte technique (Cheng et al., 2000) can be accessed commercially. The holes were made on PVC pipes of the same diameter and length. In the operational configurations, the jets were orientated inward, thus facing the centre of the FACE plot.

An automatic pressure regulator (SMC Corporation, Toyko, Japan, Mod. ITV3050, JPN) controlled the amount of CO2 released in the FACE and was operated by supplying variable voltage (0–10 VDC) that translated into a variable pressure in the pipes. The control of both on/off valves, that were used to open and close the different sectors of the octagon releasing CO2 and of the pressure controller, was obtained by means of a programmable microprocessor based control unit, specifically developed for POPFACE. The unit used a Motorola HC11 microprocessor (Nippon Motorola Ltd., Toyko, Japan) that was integrated in a modular system consisting of 8 analogue input channels, 8 analogue output channels, 8 digital I/O channels and a serial RS232 port. The unit was programmed using the language C and a routine library written in HC11 Assembler.

The POPFACE control program used the modified PID (Proportional Integral Differential) algorithm that was originally described by Lewin et al. (1992). This algorithm uses wind speed and instantaneous CO2 concentration measured every second by a cup anemometer (R.M. Young Company, Traverse City, MI, USA, Mod. 12005) and an Infra-Red gas Analyser (IRGA, PPSystem, mod. SBA-1, Hitchin, UK). These are located at the centre of the FACE plot and thus to calculate the voltage to be supplied to the pressure controller. In this way, the control program actually regulates the pressure inside the horizontal releasing pipes and therefore the amount of carbon dioxide required to obtain the set level of CO2 enrichment. The same algorithm also actuates the on/off valves to open the three sides located upwind of the FACE plot as a function of the most frequent wind direction calculated every 30 s from data sampled every second by a wind vane located at the centre of the plot.

Field performance of the POPFACE: high frequency CO2 concentration data and spectral analysis

The advection of a passive substance by a turbulent flow is a very complex and poorly understood phenomenon (Shraiman & Siggia, 2000). The basic understanding of turbulent mixing and transport is already established (Kolmogorov, 1941) but while these dimensional arguments are successful in explaining the average rate of spreading or mixing in the so called inertial subrange, understanding the large fluctuations is more a difficult problem. The nontrivial statistical aspects of the scalar turn out to originate in the mixing process itself. A well-established phenomenological parallel exists between the statistical description of mixing and fluid turbulence itself.

Spectral analysis of rapid fluctuations in CO2 concentration provides information on the time and length scales of the fluctuations, thus providing the opportunity to analyse the quality of the air-CO2 mixing. This type of information was considered critical for characterizing the CO2 environment experienced by the plants.

To perform this type of analyses, wind and CO2 data were acquired at 10 Hz using a 3D sonic anemometer (Metek GmbH, Elmshor, Germany, Mod. USA-1) connected to a fast response CO2 gas analyser (Li-Cor, Mod. 6252) with a typical sampling frequency of c. 5 Hz (Aubinet et al., 2000). The measurements were repeated, during a number of field campaigns, in the centre of the FACE plot just above the canopy height and were made both in the prototype and in the operational system. Spectra were calculated using EdiRe software (University of Edinburgh, Vs, Edinburgh, UK) from measurement periods of 30 min by calculating the Fast Fourier Transform (FFT) of de-trended fluctuations around the 30 min average.

Field performance of the POPFACE: temporal and spatial control of CO2 concentrations and CO2 use

The performance of the FACE system was evaluated by recording, throughout the experimental season, the 1-min average CO2 concentration measured by the IRGA located at the centre of the FACE plot. Such mean was calculated as average of IRGA readings made every second.

The spatial distribution of CO2 concentrations was monitored during some days throughout the experimental season using a multiport scanner that was specifically developed for the POPFACE project. The same type of microprocessor-based unit described above was programmed to operate the multiport scanner. This was equipped with 12 on/off solenoid valves and two IRGAs (PPSystem, mod. SBA-1, Hitchin, UK). The two IRGA were cross calibrated using a reference gas mixture at the beginning of each spatial concentration monitoring campaign. The samples were taken along four main octagon cords at 4 and 8 m distance from the ring centre. The actual location of the sampling points is shown in Fig. 2. The use of two instead one IRGA reduced the time period elapsing between two consecutive readings at each location. The overall CO2 use of the replicated POPFACE system was monitored by recording at 1 min interval the voltage supplied to the pressure regulator, as well as the exact amount of liquid CO2 that was used to refill the main bulk container at almost weekly intervals. Actual data were finally compared with the calculated optimal CO2 use to derive a CO2 fumigation efficiency parameter. The calculation of the optimal CO2 use was made assuming that the amount of carbon dioxide that is necessary to add 200 µmol mol−1 of CO2 to a given volume of air is a function of the air volume to be enriched and the mean air flow through such a volume. The actual amount of CO2 that was lost by vertical and lateral transport out of the enriched volume gives an estimate of the efficiency of the FACE system.

Results and Discussion

The relationship between pressure and CO2 flow rate from gas jets

The velocity of a gas jet out of a venting hole is a function of the difference between the static pressure inside the pipe (Pr) and ambient pressure (Pb). Assuming that the CO2 flow through a nozzle is isoentropic (there is no major heat exchange), the theory says that such velocity cannot be accelerated to a velocity greater than the speed of sound (John, 1984). Further increase in Pr therefore drives changes in density and mass flow out of the jet, but not in velocity. When this situation is reached, the gas jet is defined as choked. Sonic velocity depends on temperature and on the gas properties and an estimate of the Pr/Pb ratio that just chokes the nozzle can be obtained by means of the following equation:

image( Eqn 1)

(Ma, Mach number; γ, specific heat ratio (the ratio between specific heat at constant pressure and specific heat at constant volume).)

Assuming that for choked nozzles Ma is equal to 1, and that γ of CO2 is equal to 1.3, the equation yields a value of Pr/Pb of 0.547. For a standard atmospheric pressure of 101.3 KPa this implies that nozzle choking is reached when the pressure inside the pipe is approximately 0.055 MPa. This calculated value is very much in line with the experimental observations made in the laboratory, which showed that the relationship between mass flow rate and pressure was not linear when pressure dropped below 0.05 MPa. This reflected the fact that mass flow was driven, under those circumstances, by changes in both flow and velocity (Fig. 4) and not only the former.

Figure 4.

The relationship between CO2 mass flow (g min−1) and absolute pressure (MPa) measured and calculated for into a 20-mm diameter polyethylene pipe carrying one single hole of 0.3 mm diameter. Measured data and associated standard errors are indicated by solid symbols, while the open symbols are values calculated by the Computational Fluid Dynamics (CFD) model described in the text. The line was fitted by eye to the data.

These considerations were at the basis of the choice of the size of the pipes, the diameter of nozzles and the range of operating pressures for the POPFACE system. Such choice ensured that a linear relationship between pressure and flow would have been maintained for most of the time during FACE operations and that the nozzles were choked.

Air-CO2 mixing in high velocity jets: simulation and observations

The use of choked jets at sonic velocity to release CO2 has a crucial importance for the initial CO2 and air mixing. The so-called ‘shock-wave effect’ of sonic jets (Munson et al., 1990) may favour such mixing. The existence of such an effect was clearly detectable by the direct measurement of CO2 concentration near the jet and was well simulated using the CFD model. According to both the measurements and the simulations, the CO2 released from the jet was mixed very rapidly with air (Fig. 5) and the similarity between simulated and observed data also provided a substantial validation of the CFD model.

Figure 5.

Air-CO2 mixing measured and calculated by the Computational Fluid Dynamics (CFD) model downwind of an individual CO2 gas jet at sonic speed. The measurements (solid symbols) and simulations (open symbols) of CO2 concentrations are those observed/calculated along a line perpendicular to the wind direction on the same plane of the releasing jet (x-axis). The line was fitted to the data using a polynomial approximation.

The model proved critical in understanding the mechanism of rapid mixing. Detailed analysis of the air movement revealed that a large depression is created near the choked jet attracting a large amount of ambient air that, due to the highly turbulent regime of the flow, rapidly mixes with CO2 (Fig. 6). In practice, the model indicates that a 10-times dilution of the CO2 flow is already obtained at 2 mm downstream of the jet, reaching a 100-times dilution at only 30 mm (Fig. 5). Such dilution is almost independent of the absolute CO2 flow rate, as when the pressure is increased in the pipe, the higher density of the released gas enhances the shock-wave effect and the mixing is further enhanced. As a result the simulated CO2 concentration fields do not change significantly while the pressure is raised or decreased within the range necessary to induce jet choking. The results of three simulation runs made by varying the absolute pressure inside the pipe (0.05 MPa, 0.285 MPa and 0.4 MPa) clearly illustrated this effect. At these pressures the model predicted a significant change in the CO2 flow rate (1.32, 3.43 and 4.48 g min−1, respectively) and also showed that when sonic speed is attained and the jets are choked, the CO2 concentration fields are almost independent of the amount of gas released (Fig. 7).

Figure 6.

Direction and speed of the air flow calculated by the Computational Fluid Dynamics (CFD) model in the proximity of the CO2 releasing jet. The open arrows point to the direction of the air flow and the iso-surface indicate its velocity. Grey codes of velocity scale are indicated in the figure.

Figure 7.

Contour plots of CO2 concentrations simulated by the Computational Fluid Dynamics (CFD) model downwind of a CO2 releasing jet. The three frames refers to three separate simulations made by varying the absolute pressure inside the pipe. The greyscale represent the CO2 concentration, and the dotted line indicates the isoline at 5000 µmol mol−1.

Short-term CO2 fluctuations in FACE plots and spectral analysis

The main objective of a FACE system is to obtain CO2 enrichment over an open area and minimizing fluctuations in CO2 concentration. But since short-term variability associated with natural turbulence is unavoidable (Hendrey et al., 1997), the objective of the CO2 control is to keep long-term averages of CO2 concentration as close as possible to a predefined target value. There is some consensus in the scientific community that an acceptable performance of a FACE should be that of keeping CO2 averaged for 1 min measured every second, with a mean at the centre of the plot within a ±20% variation of a preset target, and for more than 80% of the time. This same objective may also be obtained with short-term CO2 fluctuations of different amplitude and length scale. Those fluctuations depend on the turbulent structure of the atmosphere and the ratio between the size of the turbulent eddies and the volume of the air where active mixing, caused by the jet, is occurring. The working hypothesis for the POPFACE design was that the volume of air in which the air-CO2 mixing is an active process caused by the energy of jets, is of the same order of magnitude as the volume of the eddies.

In our analysis we explored this relation by analysing the temporal variations of wind turbulence and CO2 concentration. The analysis gave a very good indication of the performances of the releasing systems at different wind speeds and turbulent regimes. Power spectral analysis was used to determine how the power or energy is distributed over a range of frequencies or periods and the relative contributions of oscillations with various frequencies to the total variance of the time series. The power spectra of CO2 concentrations give information on the importance of various processes controlling the CO2 mixing inside the FACE ring. Three major regions can be recognized in the power spectra (Fig. 8): the high frequency region (frequency > 2 Hz) where spurious contributions to sensor output from electronic equipment and noise dominate; the inertial subrange region (frequency range from 2 to 0.1 Hz) with the well known negative power slope of −5/3 (Kolmogorov, 1941); and longer fluctuating regions (frequencies < 0.1 Hz) with a positive slope. The relative importance of these regions changes with wind turbulence, particularly the power in the inertial sub range increases as wind speed increases (Fig. 8) denoting a mixing driven by small scale eddies. At low wind speeds, a large part of the variance is associated with the last region where CO2 oscillations that are driven by the adjusted release of CO2 and by wind gusts. Conversely, at higher wind speeds fluctuations are dominated by Kolmogorov-scale eddies.

Figure 8.

Power spectra of wind (A) and CO2 concentration (B) at different winds: 0.8 m s−1 (open circle) and 3 m s−1 (closed circle). Solid lines represent the modelled inertial subrange at low winds with the typical −2/3 power slope for the wind (A) and −5/3 power slope for the CO2 (B).

When the CO2 power spectra are compared with the corresponding spectra of main wind components (Fig. 8) a difference becomes evident both at low and high wind speeds in the low frequency region. The positive slopes of the CO2 spectra are generally steeper than the wind component spectra, denoting the action of the modified PID algorithm in smoothing the longer CO2 fluctuations. This difference gives an indication of the quality of the controlling algorithm, the greater the difference the better the system can adjust to the varying wind conditions. With this kind of analysis, the POPFACE control algorithm generally performed slightly better at lower wind speeds, than at higher wind speeds, when there seems to be less fluctuation in wind speed. However the greater turbulence at higher wind speeds masked this effect. This type of analysis is also of importance as the time scales of CO2 fluctuations may potentially affect the response of stomata and photosynthesis to elevated CO2 (Cardon et al., 1994; Hendrey et al., 1997; Hendrey, 2000).

The POPFACE field performances

Long-term performances of the POPFACE system were overall satisfactory. The system was operational from 29 June 1999 to 5 December 1999 for the entire duration of the growing season of the poplars. Over this period, the three replicated systems have been operational for more than 93% of the time, with major interruptions of fumigation due to maintenance and CO2 refilling. Other minor interruptions, always in the range of hours, were due to the failure of sampling pumps for the IRGAs.

The daily duration of fumigation decreased during the season from the initial 16 h at the end of June to the 9 h at the end of the season. The mean daily values of CO2 mole fraction calculated over the fumigation period were, for the majority of cases, within a few µmol mol−1 of the target (Fig. 9). The major exceptions were apparent in days with very high wind conditions, during which the daily mean CO2 concentrations dropped significantly (Fig. 9). This is explained by the fact that a cut off was automatically introduced when the windspeed exceeded 10 m s−1, as the flow of CO2 through the pressure regulator became insufficient. The mean overall long-term CO2 mole fractions in the centres of the three FACE plots were, respectively, 545, 544.8 and 541.7 µmol mol−1.

Figure 9.

Daily mean CO2 concentrations measured in the three POPFACE rings over the season 1999. The figures show also the mean daily wind speed measured at the centre of each individual plot, over the same period of time (dotted lines). Periods during which the strong winds caused a drop of the CO2 concentrations below target value (550 mmol mol−1) are clearly identifiable.

The wind conditions in the POPFACE site were quite regular over the season, with episodes of high wind conditions that were repeated a few times during the season. On average, the wind rotated from north-east in the morning to south-west in the afternoon with significant variations in the fraction of time from one or the other direction (Fig. 10)

Figure 10.

Frequency distribution of wind directions measured in 1999 during the period of POPFACE operations. To illustrate the mean seasonal differences in wind regimes between summer and autumn, the data have been independently processed. Solid line indicates the percent frequency distribution over the period June–September while dashed lines refer to the period October–December.

The wind climate of Tuscania had a positive effect on system performances as the air-CO2 mixing was very good especially under windy conditions. The 1-min CO2 mole fractions measured at the centre of the rings were for more than 91% of time within the ±20% deviation of the preset target of 550 µmol mol−1 and for more of 75% within ±10% deviation. There were no appreciable differences between the three replicates. Those data include periods in which the ambient CO2 concentrations were higher than the target in the early morning before sunrise, periods of high wind speed when CO2 cut off did occur and periods of maintenance and gas refilling during which the fumigation was stopped for short intervals. The frequency distribution of 1-minute CO2 averages was slightly skewed compared with the ambient values, as result of the period of high wind in which the CO2 release was insufficient to reach the preset target (Fig. 11), but such an effect was not detectable in the frequency distribution of hourly CO2 means. The standard deviation of the CO2 concentrations varied as a function of the wind speed and the hour of the day, with smaller variance for the central part of the day and high wind conditions (Fig. 12).

Figure 11.

Percent frequency distribution of CO2 mole fractions measured at the centre of the three POPFACE rings for the entire fumigation period, in 1999. Solid and dotted symbols indicate the one-minute and one-hour average concentrations, respectively.

Figure 12.

Standard deviation (SD) of CO2 mole fractions measured at the centre of the POPFACE plots, at canopy height, and averaged over one-minute. (A) Changes in standard deviation with the time of the day (B) Changes with the mean wind speed. Different symbols in frame A indicate the different rings but no distinctions are made in the frame B.

The analysis of spatial variability in CO2 concentration showed the occurrence of consistent and regular patterns depending on the wind direction. CO2 concentration was higher on the upwind side of the FACE plot than on the downwind side (Fig. 13). This pattern was repeatedly observed during the season (data not shown), but it was compensated, in the long term, by changes in the wind direction (Fig. 10). However, since the average wind speed was higher from the south-west than the north-east, it must be considered that plants on the south-west side of the plots were exposed, on average, to slightly higher concentrations than those on the opposite side.

Figure 13.

Contour plot obtained by kriging of the average CO2 concentrations (µmol mol−1) (A) and standard deviation of the means (B) measured at 12 different positions in one of the three POPFACE plots, at canopy height, during a day in the summer 1999. The figure shows the existence of a well-defined gradient in CO2 concentrations under the conditions of that particular day, with higher values on the south-east side of the plot.


Data that have been presented in this paper clearly indicate that Free Air CO2 Enrichment can be satisfactorily obtained by releasing pure CO2 instead of a prediluted air + CO2 mixture. The use of CO2 jets at sonic velocity greatly enhanced the air − CO2 mixing that occurred at the releasing points and this translated into an adequate air-CO2 mixing within the FACE area under both low and high-wind conditions.

Power spectra analysis showed the relative merits of the prediluting effects of the sonic jets and plenums when compared to subsonic CO2 release. As we hypothesized, the effect of predilution was shown by the Kolmogorov scale fluctuation.

The advantages of this type of FACE design are clear. The so called ‘blower effect’ is completely removed, the infrastructure is much lighter and less disturbing than in any other FACE system, the capital cost of the infrastructure is greatly reduced and the construction of the ‘control rings’ is no longer required.

More research is instead needed to evaluate the quality of the control under atmospheric stability and in particular at night. This aspect was not specifically considered for the POPFACE system as the fumigation was stopped during night. In this respect, the direct comparison of fast-frequency CO2 spectra for a pure-CO2 and a prediluted CO2 injection system, would be of importance to understand if predilution operated by the blowers would improve the situation significantly under those conditions (He et al., 1996).

The possibility of continuing satisfactory CO2 fumigation in a rapidly-growing poplar plantation is a big challenge ahead of us. The POPFACE project will continue over the future seasons thus providing an opportunity for further testing the quality of the CO2 control with the same free air CO2 enrichment system.


The authors wish to thank Pierpaolo Pinnacoli for continued assistance with FACE operations during the first experimental season, Beniamino Gioli for substantial help during the installation of the FACE operational set-up and Toufic El Asmar for his collaboration during the period of prototype development. The POPFACE project is funded by the Commission of the European Communities, Contract No ENV4-CT97-0657 and contributes to the GCTE (Global Change and Terrestrial Ecosystem) which is a Core Project of IGBP (International Geosphere-Biosphere Programme).