Technical Comments

Early Holocene Atmospheric CO2 Concentrations

See allHide authors and affiliations

Science  03 Dec 1999:
Vol. 286, Issue 5446, pp. 1815
DOI: 10.1126/science.286.5446.1815a

Wagner et al. (1) present a high-resolution record of leaf stomatal index (SI), from which they reconstruct the concentration of atmospheric CO2 in the early Holocene (10,070 to 9,38014C-yr BP, 11,300 to 10,600 cal yr BP) (1, 2). The resulting record shows a CO2 concentration of about 270 ppmv at 11,300 yr BP, a sharp increase to about 330 ppmv at 11,260 yr BP, and (with the exception of a dip to 300 ppmv at about 11 kyr BP) a rather constant CO2 concentration at the high level of about 330 ppmv over the next 500 years.

Wagner et al. correctly state that the CO2record from the ice core of Taylor Dome (3,4)—to date considered the most reliable and precise reconstruction of atmospheric CO2 over the Holocene—has a low temporal resolution in the discussed time interval. There are, nevertheless, at least three overlapping data points (depending on the accuracy of the synchronization) from the Taylor Dome record that show CO2 concentrations between 260 to 270 ppmv (60 to 80 ppmv lower than those suggested by Wagner et al.) (see Fig. 1). The elevated CO2concentrations of the SI-based CO2 record are not found in the ice core record.

Figure 1

Top: One sigma-calibrated age ranges for the14C control points 1, 2 and 6 as an indicator of the possible age range of the CO2 record reconstructed from stomatal frequency. The labels are the same as in Wagner et al. (1). Center and Bottom: Atmospheric CO2 concentration reconstructed from stomatal index (•) (1) and direct measurements of CO2 concentration of air enclosed in bubbles in the ice cores from Taylor Dome (⋄) (3, 4) and Vostok (□) (7, 8).

Ice-core CO2 measurements are direct measurements on air that has been enclosed in bubbles. The occurrence of artefacts in earlier ice core records mainly from Greenland drill sites [enrichment of CO2 due to chemical reactions in the ice; depletion of CO2 due to fractionation during clathrate formation (5)] can be avoided by careful sample selection. In the Taylor Dome ice core we found no indications of chemical reactions producing additional CO2, and no clathrates occur (3). This record clearly shows that an atmospheric CO2 concentration from 260 to 280 ppmv was the rule during the preindustrial Holocene, including the early Holocene. We do not want to question the SI method, but we notice that there is a fundamental discrepancy between the record of Wagner et al. and the stomata density–based CO2 record of Beerlinget al. (6), who report largely scattering proxy CO2 values from 225 to 310 ppmv between 9940 and 960014C-yr.

Wagner et al. claim that the concept of relatively stable Holocene CO2 concentrations of 270 to 280 ppmv until the Industrial Revolution is falsified by their results. We believe that this conclusion is not justified.


Early Holocene Atmospheric CO2 Concentrations

Wagner et al. (1) report early Holocene (10,070 to 9,380 14C yr BP) CO2concentrations—reconstructed from the stomatal index (SI) of fossilBetula pubescens and B. pendula leaves—of ca. 260 ppmv rising rapidly to values of around 330 to 340 ppmv, with one sample dipping to ca. 300 ppmcv. These values conflict with previous CO2 reconstructions that used stomatal density (SD) of fossil Salix herbacea leaves (2, 3), and with direct ice-core measurements from Taylor Dome (4) and other Antarctic ice cores (5). All of these data-sets are consistent with atmospheric CO2levels reconstructed or measured at ca. 260 to 280 ppmv over this time period.

Wagner et al. used a modern calibration set relating SI of leaves from one Betula pendula tree preserved over the recent (1952–1995) CO2 rise from 312 to 360 ppmv (6), supplemented by some older herbarium and field material of B. pubescens. The reconstructed CO2values appear to have been made by comparison with the regression line through the calibration set, involving its extrapolation to SI values above those in the calibration set. The reconstruction should instead have been performed by inverse regression (7) of the calibration set, where CO2 concentration is modeled as a function of SI for reconstruction purposes (8). Inverse regression has several important statistical properties (9), including minimizing the root mean squared error of prediction.

We also question whether the calibration set of Wagner et al. is of general applicability to Betula pendula andB. pubescens in the early Holocene. We have made SI measurements following Wagner et al.'s procedure onBetula pubescens leaves collected from 20 to 1100 m above sea level in southern Norway covering a pCO2 gradient of 36.5 to 31.6 Pa, and from 37 herbarium specimens collected below 500 m above sea level in Norway and Finland between 1877 and 1978, corresponding to a range of atmospheric CO2 levels of 290 to 335 ppmv. There was no statistically significant relationship of SI against CO2 in the historical data-set, suggesting that genotype, environment, and perhaps the location of the trees have a greater influence on SI than pCO2. The stomatal index in the altitudinal data-set increased significantly with altitude, but largely as a result of a slight, insignificant, decrease in SD and a large decrease in epidermal cell density (SI = SD × 100/SD + ED), such that epidermal cell density accounted for 76% of the variation in SI. These results suggest that SI varies more with individual genotype and habitat factors than with CO2concentration. A calibration set based largely on measurements from one tree growing on a peat bog may be misleading, and SI of B. pubescens and B. pendula is not sufficiently related to CO2 concentration to allow reliable CO2reconstructions from fossil leaves of mixed origin.

Wagner et al.'s conclusion that “during the early Holocene, atmospheric CO2 concentrations that were >300 ppmv could have been the rule rather than the exception” conflicts with other independent measurements or estimates of CO2concentrations (2–5). This anomaly may result from the deficiencies in their modern calibration set and the great variability in SI, SD, and epidermal cell density among Betula trees.


Early Holocene Atmospheric CO2 Concentrations

Response: The comments by Birks et al. misrepresent our interpretation of stomatal index (SI) changes inBetula leaves in terms of changes in atmospheric CO2 concentration (1, 2). A principal argument seems to be based on the observation that SI values for B. pubescens leaves collected in Norway and Finland do not match the historical relation between mean SI and CO2 as reconstructed in the Netherlands.

In order to determine the rate of historical responsiveness to atmospheric levels of CO2, it is essential to use high-resolution SI data that (i) have not been significantly affected by influences of environmental conditions other than CO2values, and (ii) that can be calibrated against the (1958–present) Mauna Loa record of increases of atmospheric CO2 levels. Because of generally low temporal resolution and insufficient environmental information, herbarium data are not necessarily reliable. For our calibration, we therefore relied primarily on the near-annual record of leaves in currently accumulating peat that can be linked to individual trees or populations (2). Through consistent harvest of modern leaves, calibration is annually updated.

The herbarium-based SI data from Norway and Finland have a low temporal resolution. The period of Mauna Loa monitoring is underrepresented. More importantly, the effect of latitude on SI values may not have been taken into consideration. A systematic study of modern Betula leaves along a north-south transect from northern Scandinavia to the Netherlands has confirmed the dependence of SI values on progressively changing photoperiods (3). Mean differences in SI values can reach 3%. In CO2 estimates, such differences would constitute a 50 ppmv discrepancy. The most sensitive geographic ranges are latitudes between 55 and 70N. Consequently, in contrast to the latitudinally restricted data from the Netherlands, herbarium data combining random localities in Scandinavia could well have been influenced by the relation between SI and photoperiod. In any event, it is obvious that response curves have to be separately established for latitudinally contrasting regions. Our ongoing studies in Finland corroborate the applicability of Betula leaves from high-latitude peat deposits in correlating SI data with increasing CO2 levels. In contrast to the suggestion of Birks et al., stomotal frequency analysis of genetically controlled material (B. pubescens, B. pendula, B. nana, and related hybrids) indicates that mean SI patterns are hardly influenced by genotypic variability (4).

The comment on statistics makes a point that may be technically jusitified, but does not invalidate any of the conclusions of our report. In our reconstruction of CO2 concentrations, we experimented with both classical and inverse regression analysis. As can be expected, there is no conflict between the two techniques when mean SI values for fossil leaves correspond to modern values that have been calibrated against the Mauna Loa record. Only the CO2estimates for our earliest data points had to be used on extrapolation, under the assumption (not yet validated by controlled-environment experiments) that the high SI values still reflect the linear part of a sigmoid stomatal frequency response to changes in atmospheric CO2 levels. Within comparable uncertainty limits, the two techniques result in a 15 to 20 ppmv difference in the estimated CO2 concentration. In order to facilitate comparisons with the Antarctic ice core measurements, we have preferred the lower estimates (260 ppmv) derived from classical regression analysis.

Both Birks et al. and Indermühle et al. note that our results conflict with previous estimates of atmospheric CO2 levels based on stomatal frequency data for Salix herbacea (5). However, apart from the fact that the applied training set includes field and herbarium material from altitudinally and latitudinally contrasting growth areas in Scandinavia, the United Kingdom, and Austria, stomatal frequency analysis of S. herbacea was restricted to SD measurements. It should be noted that SI values constitute a much more sensitive proxy for detecting changes in atmospheric CO2 levels (6). SI values express stomatal frequency changes independently of variation in epidermal cell size, and are therefore less influenced by environmental factors other than CO2levels.

We realize that our results cannot be reconciled with the ice-core records from Antarctica. Although the time-correlation is not necessarily correct, the figure composed by Indermühle et al. (figure 1) illustrates that some of the Taylor Dome data points would correspond to our Preboreal record. We have no explanation for the discrepancy. The early Holocene CO2 trend has also been detected in a record of Betula leaves from Denmark where, in addition to the Preboreal Oscillation, younger cooling events at 7.2, 3.4, and 2.5 kyr also correlate with lowered CO2concentrations (7). We believe that these preliminary results, in combination with the consistent replication of the familiar CO2 increase after the Industrial Revolution, provide ever-increasing evidence for the integrity of leaf-based CO2 quantification.


Stay Connected to Science


Navigate This Article