Climate and phenology
The Ts, cumulative precipitation, soil water content and LAI during the growing season are shown in Fig. 1. Soil temperature at the 2.5 cm depth reached a maximum around DOY 166 (Fig. 1a). The summer of 2003 was relatively dry, with only four rain events ≥ 10 mm. Total precipitation from DOY 121–274 (1 May to 31 September) was 400 mm (Fig. 1b), which was 26% less than the 30-year (1971–2000) climate normal. Soil water content remained relatively low over much of the measurement period (Fig. 1c), ranging from 0.16 to 0.34 m3 m−3, and generally remained < 0.23 m3 m−3 late in the growing season after the corn had tasselled. The corn emerged around DOY 135 (15 May) and was harvested on DOY 287 (14 October). The maximum LAI was 4.5 m2 m−2, observed on DOY 219 (8 August). Critical growth stages (defined in Ritchie, Hanway & Benson 1993) were observed as follows: (1) from emergence to nine-leaf stage (DOY 135–176); (2) from nine-leaf to tassel stage (DOY 177–206), which was the rapid vegetative growth stage; (3) silking and blister (DOY 207–225); and (4) from milk to physiological maturity (DOY 226–257).
Figure 1. Climate and phenology in 2003: (a) soil temperature at 2.5 cm depth; (b) cumulative precipitation; (c) soil water content at 10 cm depth; and (d) measured leaf area index (LAI) (circles) and simulated LAI (solid line), where the growth stages are indicated by dotted lines.
Download figure to PowerPoint
Diurnal variation of the CO2 mixing ratio and carbon isotope ratio of ambient air
Figure 2 shows the ensemble diurnal variation of half-hourly CO2 mixing ratio and the isotope ratio during typical days (DOY 210–217). The average canopy height was 2.8 m, and the corn canopy had just tasselled. The daytime CO2 mixing ratio was relatively low, in the vicinity of 350 µmol mol−1 (Fig. 2a). The CO2 mixing ratio began to increase after sunset at around 1900 h and reached a maximum at midnight. The range of daytime CO2 mixing ratio was small, typically less than 40 µmol mol−1, while the range at night was often greater than 100 µmol mol−1.
Figure 2. Ensemble diurnal variations of CO2 mixing ratio and isotope ratio during day of year (DOY) 210–217 (29 July to 5 August). (a) mixing ratios of CO2 at three heights, two above the canopy and one inside the canopy; and (b) isotope ratios of ambient air at the three heights. Daytime details are shown in the magnified inset figures.
Download figure to PowerPoint
The effects of photosynthesis and respiration on atmospheric 13CO2 abundance were also indicated in the temporal variation of the isotope ratio (Fig. 2b). During the daytime, air was enriched in 13CO2 because of photosynthetic discrimination, and the isotope ratio was relatively high, about −7.6‰. At night, the values were more negative as a result of respiration and were typically around −10.6‰. Similar diurnal patterns of CO2 mixing ratio and isotope ratio were observed by Buchmann & Ehleringer (1998) over a corn canopy.
The differences in CO2 mixing ratio and isotope ratio among the three measurement heights were most evident at night. During the day, the differences were relatively small but significant (i.e. greater than the measurement precision of the TDL, see Griffis et al. 2004b). The inlet inside the canopy (1.0 m from the ground) typically measured the lowest CO2 mixing ratio and the most negative isotope ratio. The daytime relative depletion of 13CO2 inside the canopy was most likely a result of ecosystem respiration. Light attenuation may also cause 13CO2 depletion through increasing Ci/Ca (and thus photosynthetic discrimination). However, this effect is estimated to be small for C4 plants according to Eqn 5.3 and Farquhar et al. (1989).
Canopy bulk stomatal conductance for CO2
The determination of gc is critical to flux partitioning because it is directly related to the photosynthetic computation (Eqn 5). Unfortunately, the gc estimated with the PM equation can involve relatively large uncertainties resulting from lack of energy balance closure, the generally unknown contribution of soil evaporation and errors associated with EC measurements. To limit these problems, we forced energy balance closure using the Bowen ratio obtained from the EC measurements (Twine et al. 2000), and limited our analysis to turbulent conditions when the canopy was dry (83% of the data). During the full canopy period (LAI ≥ 2), soil evaporation accounted for 3–10% of the total canopy evapotranspiration, leading to an overestimation of 1.2–12% of gc (Appendix II). The uncertainty in gc, caused by measurement errors in H, λE, wind speed (u) and water vapour pressure deficit (VPD) was estimated to be approximately 30% (see Appendix II for the details of uncertainty analysis).
Figure 3 shows the ensemble diurnal variations of gc. Maximum gc was observed during the rapid vegetative growth stage (Fig. 3a), about 0.5 mol m−2 s−1. Late in the growing season, gc decreased to less than 0.2 mol m−2 s−1 (Fig. 3c). These values are in good agreement with other studies (Kelliher et al. 1995; Steduto & Hsiao 1998). For most of the experimental period, the daily maximum gc tended to skew towards the morning, which is typical for a canopy that is water stressed (Steduto & Hsiao 1998; Kurpius et al. 2003).
Figure 3. Ensemble diurnal variation of gc during critical growth stages. Error bars represent the standard error of the mean.
Download figure to PowerPoint
Isotope ratio of ecosystem respiration
The uncertainties in the flux ratio and Keeling estimates of nightly δR using 2 min data collected over the entire night are shown in Table 1. The uncertainty of the flux ratio estimate was typically 0.7‰, which was larger than the Keeling estimate of about 0.4‰. The uncertainties of both methods increased significantly as the range of CO2 mixing ratio decreased, an occurrence that has also been reported in other studies (Pataki et al. 2003). For instance, to limit the uncertainty to within 3‰, the Keeling method required ≥ 10 µmol mol−1 range of CO2 mixing ratio (about 99% of the night-time observations met this threshold), and the flux ratio method required ≥ 10 µmol mol−1 range of CO2 mixing ratio difference between the two inlets (about 92% of the night-time observations met this threshold).
Table 1. The uncertainties in the estimation of nightly δR. The uncertainty of the Keeling estimate is the standard error of the intercept. The uncertainty of the flux ratio estimate is calculated as (standard error of the slope / Rstd) × 1000 (‰). Two-minute data collected over the entire night when u* ≥ 0.1 m s−1 were used for both methods
|Statistics||Flux ratio||Keeling intercept||Units|
|CO2 range||40a||102||µmol mol−1|
Night-time hourly δR obtained with the flux ratio method showed considerable variations, with values typically fluctuating by about 2.0‰. Bowling et al. (2003) also reported significant hourly variation of up to 6.4‰ for δR in a grassland. For a typical nocturnal pattern, hourly δR values decreased with time and reached a minimum value before sunrise. There are two possible explanations for this observation. Firstly, the isotopic composition of rhizosphere respiration may have become more depleted during the night as the substrates that were assimilated during the daytime were consumed and microbes switched to other available substrates that were less enriched. Secondly, as the night progressed, the above-ground foliar respiration could have been inhibited more than the below-ground heterotrophic respiration as a result of decreasing air temperature or substrate availability. Both mechanisms could cause δR to become relatively more depleted over the course of the night. The considerable hourly variation of night-time δR implies that there might be important limitation on the extrapolation of night-time δR to daytime values in the isotopic flux partitioning. Daytime chamber measurements on the individual component of ecosystem respiration are needed to examine this issue in further detail.
The flux ratio and the Keeling estimates of nightly δR are shown in Fig. 4. On average, the Keeling method tended to give lower values particularly during the full canopy closure period (LAI ≥ 2, DOY 177–257). One possible explanation for the discrepancy is the fetch mismatch between the two methods. A footprint analysis (Schuepp et al. 1990) showed that the effective fetch of the mixing ratio measurement at night was typically > 270 m, which was larger than the 200 m fetch of the site. It is possible that the advection of depleted CO2 from the surrounding C3 vegetation (forests and some agricultural crops) and combustion sources had influenced the mixing ratio measurements, resulting in lower Keeling estimates. In comparison, the flux measurement (derived from mixing ratio gradient) typically had an effective fetch of less than 150 m, implying that the influence of advection on the flux ratio method was smaller.
Figure 4. Seasonal variation of δR in 2003. Error bars indicate the uncertainties (standard errors) of the flux ratio and Keeling estimates, respectively.
Download figure to PowerPoint
Another possible reason for the difference is the variation of boundary layer atmospheric CO2. As CO2 is respired and mixed into the nocturnal boundary layer, the mixing ratio and the isotope ratio of ‘background CO2’ (Cbg = Ca −CR, where CR is the mixing ratio of respired CO2) varies over the course of the night. In addition, it is possible that advection from C3 vegetation or combustion sources could have affected the ‘background CO2’ considerably. This temporal variation of boundary layer CO2 introduces a bias into the night-time Keeling estimates. We hypothesized that at an hourly timescale the changes in the nocturnal boundary layer characteristics are smaller and have less influence on the Keeling estimates. We observed that the nightly Keeling estimates were generally 2‰ lower than the hourly values averaged over the entire night (hourly estimates with a standard error of σ < 1‰ were used). The flux ratio estimates did not show such significant difference between the hourly and nightly values. For hourly δR values, the flux ratio and the Keeling methods showed a typical difference of 1.7‰, which is smaller than the 3.5‰ difference for nightly values. At this point, the influence of background CO2 variation on the Keeling estimates cannot be evaluated directly. Boundary layer modelling is needed to explore this issue in greater detail. Taking into account the possible influence of advection and background CO2 on the Keeling plot and to be consistent in terms of footprint match with FN and δN, we used the flux ratio method for nightly δR in the following partitioning.
Considerable seasonal variation in night-time δR was observed (Fig. 4). During the spring (DOY < 160), values fluctuated between approximately −32 and −19‰, and then increased rapidly to a maximum of approximately −11‰ in early August (around DOY 214) before declining in the fall (> DOY 255) to values in the range of −32 to −20‰. This seasonal variation of δR was strongly related to the canopy development, as indicated by a linear relationship with LAI (r 2 = 0.64, P < 0.001). We hypothesized that the higher values during DOY 182–243 (July to August) were largely attributable to the increase in the more enriched autotrophic respiration of the C4 corn. Lai et al. (2003) has also reported impacts of phenology on seasonal variation of δR over a mixed C4–C3 grassland, where δR did not show distinct temporal pattern because of the changes in C3 and C4 contributions. The rapid decline in δR at the conclusion of the growing season was somewhat unexpected. However, during this period (DOY 255–263), a rain event increased the θ from 0.2 to 0.25 m3 m−3. Soil temperature at 2.5 cm depth also increased from 10 to > 20 °C (data not shown). The relatively warm, moist conditions likely stimulated heterotrophic respiration, causing a decrease in δR.
δR has also been shown to be influenced by precipitation, soil water content and VPD (Ehleringer & Cerling 1995; Bowling et al. 2002; Ometto et al. 2002; Fessenden & Ehleringer 2003; Lai et al. 2004; McDowell et al. 2004), which could explain some of the large day-to-day variations in δR. For instance, the abnormally low values of δR on DOY 235, 255 and 262 were preceded by rain events within 48 h. However, the low values of δR on DOY 190 and 198 were not related to θ or precipitation. Those days were characterized by low air temperature and low solar radiation, both of which would be expected to have a greater negative impact on the above-ground foliar respiration than the below-ground heterotrophic respiration. In addition, the reduced carbon assimilation, partly reflected in the low FN values, might also limit autotrophic respiration.
Isotope ratio of net ecosystem CO2 exchange and photosynthetic discrimination
Figure 5 shows the ensemble δR and δN estimated with the flux ratio method. The figure also shows Δcanopy and the isotopic ratio of the canopy photosynthesis (δP) obtained after FA was partitioned. δN depends on the magnitude of FN, δR and Δcanopy. It is therefore not surprising that relatively large fluctuations were observed in δN. For the majority of the daytime values, δN varied between −12 and −4‰. The Δcanopy remained relatively constant both diurnally and seasonally in the vicinity of 3‰, which was smaller than the value of approximately 4‰ deduced from the difference between the measured plant isotope ratio (top leaves −11.8 ± 0.4‰, bottom leaves −12.3 ± 0.3‰) and the mean ambient air ratio of −7.8‰. The relatively constant Δcanopy resulted from the positive contribution of PEP carboxylase to the assimilation of 13CO2, about −5.7‰ in terms of discrimination.
Figure 5. Ensemble diurnal variations of δR (solid line), δN (dotted line with circles), δP (dot-dashed line), δa (solid line with points) and Δcanopy (dashed line).
Download figure to PowerPoint
The Δcanopy is an important factor in some biophysical models (e.g. SiB2) that are used to partition the carbon budget at the global scale (Ciais et al. 1995; Fung et al. 1997). Fung et al. (1997) showed that a change of 3‰ in annual mean Δcanopy could cause a 0.7 Gt C year−1 bias in the carbon sink. Here, the estimate of Δcanopy is smaller than the global mean value of 3.6‰ for C4 plants reported by Lloyd & Farquhar (1994). While the Δcanopy remained relatively stable in our study, the Δcanopy of C3 or mixed C4–C3 ecosystems typically shows significant diurnal and seasonal variations resulting from changes in Ci/Ca. Lai et al. (2003) observed an apparent diurnal pattern of Δcanopy ranging from about 0.7‰ in the morning and evening to approximately 2.8‰ at midday over a mixed C4–C3 grassland. Bowling et al. (2001) also observed a strong diurnal variation of Δcanopy, approximately 16 to 19‰ over a temperate deciduous forest.
Partitioning net ecosystem CO2 exchange into photosynthesis and respiration
Half-hourly FN was partitioned into photosynthesis and respiration using both the isotopic approach and the regression method for the full canopy period. The diurnal ensemble partitioning for each growth stage is shown in Fig. 6. In general, FA ranged between −30 and −50 µmol m−2 s−1, and FR typically varied from slightly above zero to 15 µmol m−2 s−1. These results are consistent with other values in literature (Grant et al. 1989; Pattey et al. 1991; Steduto & Hsiao 1998). Both FA and FR peaked during the silking to blister stage (DOY 207–225) (Fig. 6b). The isotopic flux partitioning generally showed larger half-hourly variability than the regression method. The differences between the two methods were particularly significant at around midday, when FR sometimes showed depressions while Re continued to increase with time. No significant relationship between Re and FR was observed from 1:1 plots (figures not shown).
Figure 6. Ensemble diurnal variations of photosynthesis and respiration from the isotopic partitioning method and night-time regression method during critical growth stages. The shaded area and error bars indicate the standard error of Re and FR, respectively.
Download figure to PowerPoint
We note that the daytime pattern of Re is a direct consequence of the soil temperature and water content changes and might not represent the true physiological response of the ecosystem. Studies suggest that the night-time regression method might considerably overestimate the daytime respiration by up to 15% without considering the photoinhibition effect of photosynthesis on foliar respiration (Kok effect) (Brooks & Farquhar 1985; Janssens et al. 2001). The values of FR, however, were sometimes unrealistically low during midday (Fig. 6a & b). This might be related to the uncertainties associated with the TDL and micrometeorological measurements (discussed in further detail below). It is also important to note, however, that the isotopic flux partitioning based on mass balance principle might have considerably underestimated the photosynthesis and respiration without considering the process of CO2 recycling inside the canopy (Greaver et al. 2005).
Uncertainty analysis of isotopic flux partitioning and future recommendations
Unrealistic isotopic partitioning, that is, unrealistic values of FR, were observed, indicating that there are important uncertainties in the micrometeorological-stable isotope technique that require further investigation. For example, at around 1130 h in Fig. 6b and 1230 h in Fig. 6c, the partitioned FR was apparently below zero. In addition, the particularly high values of FR, about 15 µmol m−2 s−1, at around 1200 and 1300 h shown in Fig. 6c were unreasonable for the late growing season. These failed cases were usually associated with very low δN values, high daytime δR values, or in some cases, the unreasonably high or low gc values, assuming FN to be the ‘true’ value. Considering the magnitude of errors typically involved in the half-hourly micrometeorological measurement, which could be up to 30%, and the uncertainties in the half-hourly TDL values of δN and δR, we should not be surprised by the relatively large errors in the isotopic flux partitioning.
The total uncertainty in the partitioning of FA, propagated from errors of individual variables, was calculated according to Bevington (1969) (see Appendix II for the details of error propagation). Results showed that the total uncertainty was typically around 30% in the early growth season and increased to > 40% for the mid and late seasons when the difference between the isotope ratio of photosynthesis and respiration was small or otherwise the fluxes were small. An uncertainty test was therefore performed to examine the contribution of each variable to the total uncertainty. In order to identify the influence of isotopic disequlibrium or , we selected the early and mid seasons as two examples for which the mean δR was assigned to be −20 and −14‰, respectively. The mean values and uncertainties of Ca, gc, δN and δa were assigned constant values (Table 2). The individual contributions are shown in Table 3, which also shows the partial derivatives evaluated at the mean values. These partial derivatives indicate the sensitivity of FA to each variable.
Table 2. The means and uncertainties of the individual variables used in the uncertainty test. The uncertainties of Ca and δa are the measurement precisions of tunable diode laser (TDL) (Griffis et al. 2004b). The uncertainty of δN is the standard error of the flux ratio estimate. The uncertainty of δR is the hour-to-hour variability of night-time values
|gc||0.3||0.09||mol m−2 s−1|
|FN||−25||0b||µmol m−2 s−1|
Table 3. Contributions of individual variables to the total uncertainty in FA. pi represents the contribution of individual variable xi to the total uncertainty. The typical mean value x0 and the uncertainty of each variable are given in Table 2
|Variable xi||Early season||Mid season|
|pi (%)||pi (%)|
|Ca||< 0.01||< 0.01||< 0.01||< 0.01|
The results indicated that the uncertainties in δN estimates imposed the most significant influence on the partitioning, accounting for more than 80% of the total uncertainty (Table 3). The partitioning was also highly sensitive to δN, with a 1‰ fluctuation in δN resulting in a 10 and 25% change of FA for the early and mid season, respectively. As reported in other studies, the precise estimation of δN (or isoflux δN · FN) is essential to the isotopic flux partitioning (Bowling et al. 2001; Lai et al. 2003; Ogée et al. 2003, 2004). In this study, the uncertainty in the hourly flux ratio estimate of δN was large, ranging from 0.5 to 6‰ because of the relatively small differences in the CO2 mixing ratios between the two inlets during the daytime (< 1.5 µmol mol−1). The precision of δN could be improved by increasing the separation between the two measuring levels to obtain larger gradients in 12CO2 and 13CO2 or by extending the sampling period. Buffer volumes could be used to reduce the noise caused by turbulent fluctuations in the TDL measurements of 12CO2 and 13CO2. The uncertainty in FA could also be reduced by measuring the flux of 13CO2 directly with EC, by which the problems related to footprint mismatch and small gradient on the isoflux estimation would be overcome. For instance, the total uncertainty in FA will decrease by approximately 50% for both the early and mid growing seasons (according to the values in Table 2 and by considering δN × FN as a single variable), if the assumption of 20% error for the 13CO2 flux measurement is made.
The fluctuation in δR also contributed considerably, though much less than δN, to the total uncertainty (Table 3). The uncertainty in δR accounted for approximately 17% during the mid season when the isotopic disequlibrium was small (≈ 3‰), and about 2% early in the growing season when the disequilibrium was large (≈ 9‰). Based on the
sensitivity of the partitioning to δR[ in
Table 3], the 2‰ difference between the hourly flux ratio and the Keeling estimates of δR resulted in an approximate 4% change in FA in the early season and up to > 30% variation in the mid season when the isotopic disequilibrium was relatively small.
The gc estimate was expected to involve large uncertainties (see Appendix II). However, the contribution of gc errors to the total uncertainty varied considerably with the errors of other variables. For instance, the error in gc accounted for less than 1% of the total uncertainty when the uncertainties of δR and δN were relatively large (i.e. > 2‰). The error in gc accounted for more than 25% when the uncertainty of δR and δN was limited to within 0.5‰.
The influence of isotopic disequilibrium on flux partitioning is significant because the methodology will fail if δP equilibrates with δR. Ogée et al. (2004) showed that a small disequilibrium between δP and δR resulted in large uncertainties in the partitioning even after the precision of δR and δN was improved. Our results showed that the isotopic partitioning produced better results (indicated by less negative FR values) before DOY 185 when the difference between δP and δR was relatively large (> 5‰). According to the uncertainty analysis, the total uncertainty in FA increased from ≈ 30 to > 40% when decreased from ≈ 9‰ in the early season to ≈ 3‰ in the mid season.
Although the partitioning showed high sensitivity to δa (Table 3), the instrumental errors in δa measurements are typically small and have a small influence (less than 0.1%) on the partitioning. The uncertainties associated with parameters a, b3 and b4 require further investigation and were not explicitly considered here. Nevertheless, the partitioning appears to be highly sensitive to these parameters. For instance, if the value of b3 is changed from 27 to 29‰, the FA would be altered by about 10% on average for all seasons. In addition, it is unlikely that φ will remain constant for different varieties of corn or for different environmental conditions. The assumption of the analogy between the leaf-scale and the canopy-scale discrimination needs further investigation. Leaf-scale measurements of isotopic exchange, stomatal conductance and photosynthesis, combined with physiological modelling, could provide additional insight into these issues.