Free-air CO2 enrichment (FACE) using pure CO2 injection: system description


Author for correspondence: M. Okada Tel: +81 19 643 3462 Fax: +81 19 641 7794


  •  A free air CO2 enrichment (FACE) system in which rice was grown under elevated CO2 conditions by releasing high pressure, pure CO2 from emission tubes surrounding the crop is described here. Unlike other (FACE) systems, blowers were not used to mix the emitted CO2 with the surrounding air.
  •  Four 12-m diameter emission structures (‘rings’) were constructed. Monitoring and control of CO2 emission was carried out by a series of CO2 and wind sensors, data loggers, controllers and valves. The target CO2 concentration ([CO2]) was 200 µmol mol−1 above ambient; enrichment was carried out continuously.
  •  Temporal [CO2] control was adequate, with c. 60 and 90% of the air samples at ring center having a [CO2] within 10 and 20% of the target, respectively. Spatial [CO2] distribution was also adequate, with 60% of the ring area having a [CO2] that was within 15% of that at the center.
  •  At comparable wind speeds, the pure CO2 injection FACE system described here had a similar performance to that of FACE designs that use blowers to mix the injected CO2 with the air.


The anticipated increases in global atmospheric CO2 concentrations ([CO2]) have led to much research that aims to determine the growth of plants when exposed to the levels of [CO2] predicted for the latter part of this century. The free-air CO2 enrichment (FACE) technique has been successfully used to grow a wide variety of vegetation types under elevated [CO2] (for example, wheat (Kimball et al., 1995), pastures (Hebeisen et al., 1997) and trees (Hendrey et al., 1999)). Compared with methods that grow plants under elevated [CO2] in enclosures, the main advantage of FACE is that it does not substantially modify environmental factors such as incident solar radiation, temperature, humidity and wind, which can influence the response to elevated [CO2] (see McLeod & Long, 1999). Generally, FACE systems enrich a circular area of vegetation with CO2 in order to generate a zone with a higher [CO2] than that of the surrounding ambient atmosphere. The CO2 is usually emitted from a structure (commonly referred to as a ring) constructed from pipes or tubes that surrounds the vegetation and is dispersed across the vegetation by the wind.

One major problem with many existing FACE designs is that they use blowers or fans to predilute and evenly mix the injected CO2 with the air. Under stable wind conditions, the blowers can impart small but potentially important effects on the microclimate inside the ring. For example, under night time inversion conditions at the USDA Water Conservation Laboratory Maricopa FACE site (Arizona, USA), which uses the Brookhaven National Laboratory (BNL) enrichment system, the blowers draw down the warmer air higher in the inversion, leading to canopy air and foliage temperatures that can be up to 1°C higher than ambient plots not equipped with blowers (Pinter et al., 2000). Since blower-induced canopy temperature increases usually occured at night when inversion conditions and low wind speeds are common, the blower phenomenon has been used as an argument to stop CO2 enrichment at night. However, since elevated [CO2] may influence respiration rates (Drake et al., 1997), night time enrichment may be necessary to reflect the effects of elevated [CO2] on crops accurately. Therefore, in order to release CO2 at night and to account for the possible effects of blowers, a FACE experiment must have ‘blower only’ control plots. Also, some FACE experiments have an additional set of plots with no blowers so that any possible effects of the blowers only on crop growth can be accounted for. Such an experimental set up will not only increase installation and running costs but also necessitates an extra set of measurements and samplings.

To avoid any potential blower problems, we designed a FACE system that does not rely on blowers to mix the CO2. Instead, pure CO2 is emitted at pressure through tubes with numerous tiny holes, resulting in a quick and efficient mixing of CO2 with the surrounding air. The mixed CO2 is then dispersed across the ring by the wind. Because our FACE design was developed to test the effects of elevated [CO2] on rice, it has other features that make it different from other FACE designs. Rice is unique among the major agricultural crops in that it is usually grown under flooded conditions, making for muddy and soft support surfaces during the growing season. As such, the CO2 emission structures of the Rice FACE project are of light construction, which also greatly facilitates their installation and dismantling.

Because a large proportion of the world’s population depends on rice for a significant part of their dietary needs, it is important to determine the effects of the predicted increases in atmospheric [CO2] on rice growth and yields. The Rice FACE project, the first experiment to grow rice under elevated [CO2] without using enclosures, was instigated in 1996, with design trials being carried out in 1997. The full-scale project was set up in 1998 with a second year of enrichment being carried out in 1999. Here we describe in detail the design and construction of the FACE rings and present system performance data from the 1999 season.

Materials and Methods

System description


The Rice FACE experiment is located in Shizukuishi town, Iwate prefecture, in the northern part of Honshu, Japan (39° 38′ °N, 14057′ E). The area is typical of the agro-environment that grows a large proportion of the Japanese rice crop. It is situated at an altitude of about 200 m and has a ‘humid continental’ climate with a summer precipitation maximum and a cold, dry winter. Daily average air temperatures range from −2.5° (January) to 23.2°C (August) and precipitation averages about 1200 mm per annum. To establish the crops for the FACE experiment, agronomic techniques typical of the local area were used. A full description of the crop history is provided by Kim et al. (2001). Briefly, seedlings of rice cv. Akitakomachi were transplanted in late May into conventionally prepared paddies. Leaf area index peaked at about 5 in mid-August. Final harvest at grain maturity was in late September. Plants in the FACE plots produced about 15% more total dry matter (14 t ha−1) and grain yield (6.5 t ha−1) than ambient plots (see Kim et al., 2001).

System layout

The Rice FACE facility consisted of four octagonal CO2 enrichment rings (hereafter referred to as FACE rings) together with their four companion ambient (nonenrichment) plots (Fig. 1). The FACE rings and ambient plots were referred to as A to D and W to Z, respectively. Each ring and its ambient plot was placed in one of four blocks consisting of paddies with similar agronomic histories and soil characteristics. The ambient plots were situated at least 90 m (center to center) from any FACE rings to minimize contamination by CO2 released from the rings. Each paddy was 100 m long by 30 m wide surrounded by 10–20-cm high flood water retaining banks.

Figure 1.

Layout of the Rice FACE site showing location of rings and ambient plots, access tracks, header tanks and gaseous CO2 supply lines (solid black line).

FACE ring assembly

Each FACE ring was made of eight 5-m long flexible black emission tubes with a diameter of 38 mm arranged to make an octagon (Fig. 2). Each tube was supported horizontally at 50–60 cm above the canopy by a 5.2-m long, 22-mm diameter, galvanized steel pipe that was held up at each end by similarly sized, vertical pipes dug 40 cm into the soil. Vertical pipes were also placed halfway along each horizontal pipe for additional support. The horizontal pipes holding the emission tubes were connected to the vertical pipes with movable joints, which enabled the tubes to be easily lifted according to plant growth. The octagonal FACE ring measured 12 m across and, leaving a 1-m buffer zone from the emission tubes, had a nominal usable area of c. 80 m2.

Figure 2.

Plan view of a ring showing access walkways, control and valves housings, PVC CO2 supply tubing, CO2 emission tubes and the location of the wind sensors and air sampling points. CO2 emission tube (solid line); PVC tubing (dotted line); air sampling locations for spatial CO2 monitoring (dotted circles); centre air sampling for CO2 control and monitoring (solid circle)

The emission tube used is commercially available irrigation tubing (Kiriko type-R, Mitsui Kagaku Platech Co. LTD, Japan) made from polyethylene with a thickness of 0.5 mm. The tube is designed to be usable at a pressure of 200 kPa and has numerous tiny holes (0.5–0.9 mm in diameter) on one side. A single series of holes consisted of 12 holes differing in size spaced at 4-cm intervals in a zigzag pattern with the same pattern being repeated every 48 cm. The tubes were turned so that the holes faced into the ring center. A preliminary experiment conducted in a wind tunnel showed rapid mixing of the emitted pure CO2 gas with ambient air. The [CO2] dropped to 2000 µmol mol−1 at a distance of 20 cm from the tube surface at wind speeds of 0.2 m s−1. The relationship between the emission rate of CO2 from the tube and gas pressure is shown in Fig. 3.

Figure 3.

The relationship between CO2 emission rate (litres per meter of tube per minute) and tube gas pressure.

Soon after fertilizer had been applied and cultivation finished (mid-May), the support pipes and emission tubes were quickly put in place. On the day of seedling transplanting (20 May 1999), the CO2 delivery lines and assorted sampling tubes (see below) were carefully installed around the seedlings and emission of CO2 commenced.

CO2 supply

Liquid CO2 was stored in two vacuum insulated tanks with capacities of 10 000 kg and 15 000 kg. Liquid CO2 was vaporized through an electric heat exchanger, and gaseous CO2 at a pressure of 700 kPa was delivered through rigid PVC piping to each FACE ring site. The gas pressure was dropped to approx. 250 kPa with a pressure regulator before the gas entered to a flow control valve. The flow-rate regulated gas was supplied to an 8-port header, and each port was equipped with an electrically actuated on/off valve. The on/off valves were connected independently to the respective emission tubes of the FACE ring through a flexible PVC tube. As a result of the resistance in this supply line, when the flow control valve was fully open the maximum CO2 gas pressure at the emission tubes dropped to approx. 100 kPa. Using this pressure extended the effective working life of the tubes.

Monitoring and data acquisition

Wind monitoring

The algorithm controlling the [CO2] in the FACE rings required both wind direction and wind speed signals. At each ring center a cup anemometer and wind vane set (Model 03001–5 Wind Sentry, R.M. Young Co., USA) was placed at a height of 2.5 m above the ground to sample the wind speed and direction at 1-s intervals (Fig. 2). Since precise measurement of wind speed was not of major importance in the control algorithm, only readings above 0.3 m s−1 were used. When readings were below 0.3 m s−1, it was considered that there was no wind. When wind speed exceeded 0.3 m s−1, the wind direction signal was used to determine which set of three adjacent emission tubes were in the upwind position for CO2 release (see below).

[CO2] monitoring

There were two different configurations for monitoring [CO2] within the FACE rings; one engaged in controlling the [CO2] at the center (hereafter referred as to CTRL) and the other in evaluating the spatial [CO2] distribution (hereafter referred as to SPTL). The CTRL unit continuously monitored [CO2] at the FACE ring center at canopy height, while the SPTL unit scanned locations intermittently inside the ring at canopy height and at different positions inside the canopy. For control purposes, a SPTL unit monitored the ambient [CO2] at canopy height at the center of ambient plots W and Z (located at the northern- and southern-most ends of the FACE site; Fig. 1). Because the predominant wind direction at the site was northerly or southerly, these two plots were less likely than the other two ambient plots to be contaminated by the FACE CO2. The [CO2] inside the canopy was also determined in these ambient plots.

Both of the CTRL and the SPTL units consisted of a gas sampler and an infrared CO2 analyzer. In the CTRL unit, air was sampled at the ring center at canopy height, and the sample’s [CO2] was monitored every 1 s and averaged over 5 s for control purposes (see below). The intention was that the 1-min average [CO2] be recorded but because of an apparent error in the data logger programming language/manual, only the last 5-s average [CO2] was saved. Hence, [CO2] values were recorded for only 1/12th of the operating period. Measurements conducted for 37 d from July to August 2000 showed that, compared with using all the 5-s interval averages to calculate daily mean [CO2], using only the final 5-s sample recorded every minute (as in 1999) underestimated [CO2] values by an average of only 1.1 µmol mol−1 (SD = 1.6 µmol mol−1; maximum difference of 4.8 µmol mol−1).

The SPTL gas sampler was of a similar configuration as that of the CTRL sampler except that it also had a scanner which was used in switching sampling channels in a sequential order. The FACE ring SPTL unit was able to scan 16 different locations at 1.5-min intervals. In the 1.5-min interval, the first 1 min was used for purging and the following 30 s for measuring [CO2] every 1 s. For each location, the average [CO2] over the 30 s of each sampling interval was recorded every 24 min. The sampling air inlets were located at canopy height at ring center and at locations set equidistantly in two concentric circles 2.5 m (four locations) and 5 m (eight locations) from the center (Fig. 2). The three other sampling ports were used to sample air from the inside canopy (at a middle of the canopy height, for example 30–50 cm below the top of the canopy) at the center and at 2.5 and 5 m from the center. The ambient SPTL sampler air inlets were located at the center at canopy height, at 30 cm above the top of the canopy and at a mid-canopy height as similar as to the FACE plot. After purging for 1 min, the 1-min average [CO2] was recorded every 6 min.

A temperature compensated CO2 analyzer LI-6252 (Li-Cor Inc., USA) with a pressure transducer was used in both the CTRL and the SPTL units to measure [CO2]. The analyzers were manually calibrated every other week using zero and standard (c. 950 µmol mol−1) CO2 gas.

Data acquisition and control

For both CTRL and the SPTL operations, data loggers (CR10X, Campbell Scientific Inc., USA) were used for data acquisition and control. Power supply to the loggers and their associated devices was from a solar panel and a rechargeable battery in order to reduce the possibility of damage by lightning.

Control algorithm and software

Our objective was to maintain the FACE ring [CO2] at 200 µmol mol−1 CO2 above that of ambient. The control computer accessed the data loggers at W and Z every 6 min to collect the latest values of ambient [CO2] at the canopy top. The two values were compared and the lower one was selected and sent to all the FACE data loggers as the reference ambient [CO2] for the next 6 min. The FACE data loggers then controlled the rate and direction of CO2 emission in order to maintain the [CO2] at the center at 200 µmol mol−1 above that of ambient according to the control algorithm (see below). Wind and CO2 data were collected every 1 s, averaged over 5 s and sent to the CTRL unit. When the wind speed was over 0.3 m s−1, CO2 was released from the most upwind emission tube and the two adjacent tubes. When there was no wind (wind speed < 0.3 m s−1), CO2 was emitted from every other tube (four tubes in total). As long as the ‘no wind’ condition continued, the two sets of the four tubes were alternately actuated every 10 s.

The amount of CO2 released into each of the rings was regulated with a proportional solenoid valve (DN6, Herion-Werke KG, Germany), which was able to control a gas flow rate in proportion to the DC (0–5 V) signal input. A PID-type algorithm was used to determine valve aperture. The required change in the DC signal input was calculated from the proportional, integral and derivative components of the deviation of the FACE [CO2] from the target:

inline image

V, the change in DC signal input; Δ[CO2]n, the deviation of [CO2] from the target at present; Δ[CO2]n−1, the deviation of [CO2] from the target 5 s before; and kp, ki and kd, the coefficients of proportional, integral and derivative components, respectively.) The time of integration was empirically determined and was 1 min. Shorter integration times often led to a consistent shift in the deviation of [CO2], while longer integration times led to a slow response to fluctuating conditions. The PID coefficients were also empirically determined and the same values were used throughout the whole season. Because large and rapid changes in valve aperture often resulted in large fluctuations in [CO2], upper and lower limits were set for the DC signal input to prevent such changes. The limits were a linear function of the 5-min averages of wind speed.

Results and Discussion

System performance

FACE system performance can be reported as an assessment of the temporal and spatial control of [CO2] within the FACE rings and also as the overall reliability of the FACE system in supplying CO2. Because the Rice FACE system used a dynamic target based on the real-time ambient [CO2], control performance at any point in time was expressed as follows:

inline image

(‘FACE’, the [CO2] of a sample from a FACE ring; and ‘set point’, the ambient [CO2] plus 200 µmol mol−1 at that time.) The control set point was updated every 6 min, so to match all the CTRL time records, a data set of ambient [CO2] at 1-min intervals was created by linear interpolation between the recorded 6-min ambient values. For a given time period (e.g. for any given day or over the whole season) or environmental condition (e.g. a certain wind speed class), the fraction of recorded [CO2] values that were within 10 or 20% of the target can be calculated (that is 0.9 ≤ TAR ≤ 1.1 or 0.8 ≤ TAR ≤ 1.2). These ‘fraction limits’ can be calculated using either the 5-s averages recorded every 1 min by the CTRL logger or the 30-s averages recorded every 24 min by the SPTL logger at ring center.

Wind characteristics

Because wind is an important variable in determining the distribution of CO2 across FACE rings, information on its characteristics is essential in assessing CO2 control performance. Daily mean wind speeds ranged from less than 0.1 m s−1 to 3.5 m s−1, while the mean wind speed for the whole season across all rings was 0.79 m s−1. There were differences in the wind characteristics of the rings, with ring D experiencing less wind (mean wind speed 0.68 m s−1) than the other rings. Table 1 shows the wind speed class distribution for ring B as representative of the Rice FACE site. Wind speeds were below the detection limit (0.3 m s−1) 49% of the time, while they were above 1 m s−1 only 35% of the time (Table 1). The proportion of occasions with a low average wind speed increased as the season progressed (data not shown). Daytime wind speeds were greater than at night (Fig. 4), with the latter having frequent long, still periods.

Table 1.  Seasonal wind class distribution of ring B, Rice FACE site, Shizukuishi, Iwate, Japan. Average wind speed over the whole season was 0.79 m s−1
Wind classOccurrence (%)
< 0.3 m s−148.8
0.3−1 m s−116.5
1−2 m s−122.3
2−3 m s-1  7.1
> 3 m s-1  5.2
Figure 4.

Daily record of 5-s average [CO2] (solid circles; recorded every 1 min), target and target ± 10%[CO2] value (grey and black lines, respectively) and 10 minute average wind speed (grey bars) of ring B on August 12th 1999.

Long-term temporal control (season)

Based on the season-long mean ambient [CO2], the long-term target [CO2] for all rings was 592.6 µmol mol−1. The actual average seasonal [CO2] at ring center calculated from all the 5-s samples (including when the system was using the wrong set point; see below) ranged from 594.7 µmol mol−1 to 600.2 µmol mol−1 (Table 2). About 50 and 80% of the 5-s averages (recorded every min) were within 10 and 20% of the target [CO2], respectively (Table 2). For the 30-s averages (recorded every 24 min), 60 and 90% of the samples were within 10 and 20% the target, respectively. These performance values are lower than those reported for the Maricopa BNL-FACE facility where instantaneous ‘grab’ samples at ring center were within 10% of the target c. 65% of the time and 90% of the 1 min average samples within 10% of the target (Nagy et al., 1992). Similarly, for the Italian Mid-FACE system (560 µmol mol−1 treatment), 74% of the 1-min averages were within 10% of the target (Miglietta et al., 1997).

Table 2.  Seasonal average [CO2] at ring center, overall target achievement ratio (TAR; calculated as FACE/set point) and the fraction of 5-s samples (taken every 1 min) and 30-s samples (every 24 min) within ±10 and 20% of the target (0.9 ≤ TAR ≤ 1.1 or 0.8 ≤ TAR ≤ 1.2) for the four Rice FACE rings. Average seasonal ambient [CO2] was 392.6 µmol mol−1
Ring [CO2] (µmol mol−1)5-s sample30-s sample
TARLimit fractionTARLimit fraction
±10% ±20% ±10% ±20%

Wind had a strong influence on [CO2] control. Data for ring B is shown as representative of all rings (Fig. 5). When wind was not detected (< 0.3 m s−1), the [CO2] of the 30-s samples was within 10 and 20% of the target only about 55 and 85% of the time, respectively. This wind condition occured nearly 50% of the time over the season and may be the main reason for the difference in [CO2] control performance between the Rice FACE and other FACE systems. With a mean seasonal wind speed of only 0.8 m s−1 in 1999, the Rice FACE site had markedly lower wind speeds than the BNL-Maricopa (1.6–1.7 m s−1) and Italian Mid-FACE (1.2 m s−1) sites (Fig. 5). At wind speeds of up to about 2 m s−1, the performance of the Rice FACE system improved markedly and performance was similar to that of the other FACE systems (Fig. 5).

Figure 5.

Fraction of time that the [CO2] (30-s average collected every 24 min) was within 10 (open squares) and 20% (solid squares) of the target at different wind speeds over the whole season. Data for speeds over 4 m s−1 not shown because of the small number of data points. Mean seasonal wind speed of the Rice FACE, Italian Mid-FACE and BNL-Maricopa FACE sites also shown. Arrows indicate the mean fraction of time over the season that the [CO2] was within 10% of the target for each FACE design.

Short-term temporal control (daily)

Control of [CO2] on a daily basis is shown in Fig. 4 using data from ring B on August 12, 1999. The target [CO2] ranged from about 530 µmol mol−1 in the early afternoon to nearly 750 µmol mol−1 at night. The mean TAR for the 5-s samples was 1.02, though minute to minute values ranged from 0.65 to 2.00. Sixty and 88% of the 5-s samples were within 10 and 20% of the target, respectively.

The hourly mean TAR calculated from the 5-s samples is mapped for all days of the season in Fig. 6. At the beginning of the season control of [CO2] was unstable, showing repeated over-enrichment and under-enrichment throughout the day and night. This may have been caused by weak vertical mixing of the air (that is less turbulence) as a result of the relatively smooth surface of the paddy with only young, small plants. As the plant canopy developed over the latter half of the season, [CO2] control tended to undershoot during the day and to overshoot at night. The undershooting was observed mainly under windy conditions (when the supply of CO2 was insufficient) and the overshooting under windless conditions. As a result, the seasonal average night time TAR was higher than that during the day (1.05 vs. 1.00) largely because of a greater occurence of [CO2]‘spikes’ at night, probably due to the near absence of wind. Two short periods of all-day over-enrichment were observed between 30 and 50 d after transplanting. They occured just before the CO2 emission tube height was raised: there was a tendency for the air to be over-enriched with CO2 when there was only a small distance between the emission tubes and the top of the canopy.

Figure 6.

Hourly mean target achievement ratio (TAR) for all days of the season calculated from the 5-s samples for ring B. White bars, < 0.95; grey bars, 0.95–1.05; black bars, > 1.05.

Spatial control

Horizontal spatial distribution of CO2 was determined by assessing [CO2] at 13 locations within each ring at canopy height. Figure 7 shows the interpolated [CO2] isoconcentrations for ring A over the whole season. Across all wind speed classes, within 2.5 m of the center, [CO2] averaged 228 µmol mol−1 above ambient (Fig. 7a), and about 50% of the 30-s averages were within 10% of the target (data not shown). At 5 m from the center, [CO2] averaged 288 µmol mol−1 above ambient, and only c. 30% of the 30-s samples were within 10% of the target. The [CO2] levels at 2.5 and 5 m from the center were 3.5 and 13% greater, respectively, than at the center, resulting in a ‘bowl-shaped’[CO2] distribution pattern. Based on the interpolated values in Fig. 7(a), it can be calculated that over the whole season and averages across all rings, the central 60% of the ring’s area experienced an average [CO2] no greater than 15% above the target.

Figure 7.

Spatial distribution of [CO2] for ring A from May 25th to September 9th 1999: (a) all wind speeds; (b) wind speeds < 0.3 m s−1; (c) wind speeds > 0.3 m s−1. Interpolated data based on [CO2] values at the 13 sample locations (Fig. 2).

Wind speed had a large effect on horizontal spatial [CO2] distribution. When wind speeds were low or not detected (< 0.3 m s−1), an exaggerated bowl-shaped distribution pattern was readily apparent, with the difference between the center and peripheral [CO2] being as great as 120 µmol mol−1 (Fig. 7b). At low wind speeds CO2 was released from sets of four alternating tubes at 10-s intervals and then dispersed and diluted as it traveled towards the center. As long as the direction of these low-speed winds was evenly distributed over the season, such a distribution pattern is to be expected. By contrast to the low wind speed situation, when wind speeds were greater than 0.3 m s−1 distribution was very even, with a maximum difference between the center and edges of only 40 µmol mol−1 (Fig. 7c). Also, Fig. 7(c) shows greater differences in [CO2] across the east–west axis of the ring compared with that across the north–south axis. At wind speeds above 0.3 m s−1, the southerly wind was dominant at the site (frequency of occurrence: 44% for S compared with 22% for N, 12% for E and 22% for W). When the wind direction is perpendicular to any given side of the ring, due to an ‘angle of attack’ effect, the central emission tube will be emitting less CO2 per unit of CO2-enriched area than the two adjacent tubes, resulting in over-enrichment at the edges of the ring. Such angular effects occurred in the octagon-shaped Rice FACE rings and may have resulted in the greater [CO2] difference across the east–west axis compared with across the north–south axis.

To assess vertical CO2 distribution, in addition to the SPTL [CO2] measurements made just above the canopy, [CO2] was measured 30–50 cm below the top of the canopy at the center in both the FACE rings and ambient plots. Also, for the FACE rings measurements were made inside the canopy at locations 2.5 and 5 m from the center. The difference in the [CO2] values inside and above the canopy (diffCO2) were calculated using the inside canopy [CO2] values as a reference. During daylight hours, at the center of the FACE rings, diffCO2 averaged −19 µmol mol−1 (that is [CO2] was lower inside the canopy), while at 5 m it was −26 µmol mol−1 (Table 3). For the ambient plots diffCO2 was −17 µmol mol−1. At night, diffCO2 in the FACE rings averaged +25 µmol mol−1 at the center and was nearly +70 µmol mol−1 at the edges (that is [CO2] was higher inside the canopy). By contrast, for ambient plots night time diffCO2 was only +8 µmol mol−1. A lack of air mixing due to thermal inversion and lower wind speeds at night probably account for this large difference in the diffCO2 of the FACE rings and ambient plots. This is of concern because of the possible effects of [CO2] on night time respiration rates. If the high night time [CO2] levels inside the canopies of crops within the FACE rings (relative to the [CO2] above the canopy) affect the level of respiration, the measured effects of elevated CO2 on crop growth and yield in this experiment may not accurately reflect the responses under future levels of elevated atmospheric CO2.

Table 3.  The difference between [CO2] values measured at 30–50 cm below the top of the canopy and those just above the canopy calculated using the inside canopy [CO2] values as a reference (i.e. negative numbers indicate a lower [CO2] inside the canopy). Mean [CO2] of all four FACE rings at the center and at 2.5 and 5 m from the center together with the average of the two ambient plots are shown. Measurements taken from July 15 to August 20 1999; day samples taken from 0930 to 1330 local time and night samples from 2330 to 0330 local time
FACECenter −19.8 +24.9
 2.5 m −24.5 +29.8
 5 m −26.2 +68.6
Ambient  −17.3   +7.6

System reliability

The reliability of a FACE system in supplying CO2 at the right time and in correct amounts is dependent on all the components functioning properly within their design limitations. Malfunctions can affect both the hardware (e.g. CO2 supply lines, sampling tubes and valves) and software (e.g. control programs) components of the system. In 1999 the main problems encountered were electrical ‘noise’ in the network system and the control computer locking up. In both cases the disruption usually resulted in the wrong CO2 set point being used; this did not actually stop the release of CO2 but did result in under or over-enrichment. Post-hoc analysis showed that over the whole season, CTRL set points different from the real set point (that is ambient [CO2] plus 200 µmol mol−1 CO2) by more than 5 µmol mol−1 CO2 were utilized about 5% of time (data not shown).

CO2 requirements

Across the whole season, CO2 requirements to maintain enrichment averaged 23 kg CO2 ring−1 h−1. CO2 supply requirements for FACE systems are related to wind speed and atmospheric stability, with the latter being influenced by air temperature and solar radiation (Nagy et al., 1992). The influence of atmospheric stability on CO2 use tends to decrease with increasing wind speed. For the Rice FACE rings, in terms of both the slope of the regression line and the r2-value, there were only weak correlations between the amounts of CO2 emitted per day and average daily wind speed (Fig. 8) or levels of solar radiation (data not shown). This was most likely due to the relatively high occurence of periods with low or no wind; these occured on days with a range of atmospheric conditions (as indicated by the daily solar radiation; data not shown). As shown in Fig. 7(b), the outer volume of air was markedly over-enriched with CO2 compared with the inner volume under low or no wind conditions. Consequently, the mean [CO2] over the whole area of the ring decreased as the wind speed increased, even though [CO2] at ring center was maintained at a similar level. This over-enrichment in terms of the spatial mean under low wind conditions may be a reason for the lack of correlation between CO2 requirements and wind speed. Also, at wind speeds below 0.3 m s−1 the emission of CO2 was from the two sets of four alternate tubes every 10 s: though this provided adequate control in these conditions, it is not very efficient in terms of CO2 usage.

Figure 8.

Daily CO2 requirements per ring as a function of average daily wind speed in the 1999 season, Rice FACE site, Shizukuishi, Iwate, Japan. Each data point is the average of the four FACE rings.

The amount of CO2 emitted per unit enriched area in the Rice FACE rings averaged 0.28 kg CO2 m−2 h−1. This was a similar level of CO2 required by the BNL-Maricopa FACE system (Nagy et al., 1992) and significantly more economical than the 1.3 kg CO2 m−2 h−1 reported for the Italian Mid-FACE project (Miglietta et al., 1997).


Financial support for the Rice FACE project was provided by the CREST research program of the Japan Science and Technology Agency. We thank Richard Norby, Franco Miglietta and one anonymous referee for valuable suggestions to improve the manuscript. Many thanks to the staff, especially to Hisaya Tanaka, Rikiya Kimura and Yoshiyuki Fujisawa, of the Field Management section at the Tohoku National Agricultural Experiment Station for technical support and to Matthew W. Mitchell at North Carolina State University, USA, for help on CO2 data analysis.