A Deconvolution of the Tree Ring Based {(cid:127)13C Record

We assumed that the tree-ring-based (cid:127)3C/(cid:127)2C record constructed by Freyer and Belacy (1983) to be representative of the fossil fuel and forest-soil induced (cid:127)3C/(cid:127)2C change for atmospheric CO2. Through the use of a modification of the Oeschger et al. ocean model, we have computed the contribution of the combustion of coal, oil, and natural gas to this observed (cid:127)3C/(cid:127)2C change. A large residual remains when the tree-ring-based record is corrected for the contribution of fossil fuel CO2. A deconvolution was performed on this residual to determine the time history and magnitude of the forest-soil reservoir changes over the past 150 years. Several important conclusions were reached. (1) The magnitude of the integrated CO2 input from these sources was about 1.6 times that from fossil fuels. (2) The forest-soil contribution reached a broad maximum centered at about 1900. (3) Over the 2 decade period covered by the Mauna Loa atmospheric CO2 content record, the input from forests and soils was about 30% that from fossil fuels. (4) The (cid:127)3C/(cid:127)2C trend over the last 20 years was dominated by the input of fossil fuel CO2. (5) The forest-soil release did not contribute significantly to the secular increase in atmospheric CO2 observed over the last 20 years. (6) The pre-1850 atmospheric Pco2 values must have been in the range 245 to 270 x 10 -6 atmospheres. to be given regarding the magnitude of the forest-soil contribution to the build-up of atmospheric CO2.


INTRODUCTION
With the publication of a paper by Bolin [1977] a controversy began regarding the contribution of the terrestrial biosphere and soils to the rise in atmospheric CO2 content observed over the last 20 years. Bolin pointed out that the CO2 released as a result of deforestation and of the agricultural manipulation of soils could not be taken as negligible when compared with CO2 released through the combustion of fossil fuels. Other investigators followed uP this idea (see summary in Table 1). While most agree with Bolin's basic conclusion, a few have gone much further and proposed that the forest-soil contribution over the last 20 years has equaled or even exceeded that from fossil fuels. Geochemists, interested in fossil fuel CO2 uptake by the sea, have, on the other hand, generally been critical of the contemporary forest-soil contribution estimates because their models fall somewhat short of explaining the difference between the amount of CO2 produced by fossil fuel burning since 1958 and the amount of excess CO2 accumulated in the atmosphere since 1958. Hence, they see no place to store the CO2 coming from forests and soils . They are sufficiently confident in their models to preclude entirely those scenarios that involve large forest-soil releases. Only if pushed to the limits of their credibility could these models accommodate even the modest amounts of forest-soil CO: suggested by Bolin. When this debate arose, the most promising approach to its resolution appeared to be through the 13C/12C record for atmospheric CO2 contained in tree rings. Stuiver [1978] was the first to derive from such a record the magnitude of historical forest-soil atmospheric release. As summarized in Table 1, his estimates of the rate and total magnitude of the release were comparable with those of Bolin. Subsequent •3C work, although based on a larger data base and interpreted by using more rigorous models, has not significantly In addition to those authors claiming success in obtaining a meaningful 13C record (see Table 1), there are quite a number who have obtained records divergent to the expected trend. We have attempted to summarize all this work in Table 2 and in Figure 1. A quick look at these summaries certainly does not raise one's confidence in the 13C/12C approach. Clearly, the changing atmospheric 13C/12C ratio cannot be the only source of variability in the carbon isotope ratios for the wood formed in a single tree. Other factors  Bolin [1977] 0.8 _+ 0.5? 5.8 _+ 2.5 (1800-1975)? Adams et al. [1977] 4.2? --Stuiver [1978] 1.05 10 (1850-1950)$ Woodwell et al. [1978] 8.3 -+ 6.7? -- Wong [1978] 0.8? -- Wagener  With this in mind, we ranked ( Figure 1) each of the existing records for material (M), site (S), and climate (C). A cross indicates that the record is deficient with regard to this criterion. On the basis of these criteria, only nine (reference numbers 5, 6, 7, 15, 20, 21, 22, 23, and 24) fully qualify. In constructing his composite curve, Freyer used seven of these records (5, 6, 7, 20, 21, 23, and 24) and 11 of those not fulfilling the full set of criteria (4, 8, 16, 17, 18, 19, 26, 29, 30, 31, and 33). In Figure 3 we compare Freyer's composite with that based only on the nine records fulfilling all the criteria set forth above. For those records in Figure 1 that include the juvenile stages of tree growth, a systematic increase in 13C with time is often observed. This 'juvenile effect' has been attributed to either reassimilation of respired CO2 that has accumulated beneath the forest canopy [Freyer, 1979a] or to reduced CO2 assimilation rates due to lower irradiance near the forest floor [Francey and Farquhar, 1982]. To eliminate the inclusion of juvenile effects in the composite, the first 20-30 years of growth for each tree were excluded during averaging. The specific deletions are noted in the caption to Figure 3.
As can be seen, the 2-decade-averaged curves, although disagreeing in detail, show the same linear decline from 1850 to 1970. Also shown in this figure is a comparison between the Freyer composite and trees selected only for material and site (i.e., the climate criterion is dropped). In this case, most of the same records are used, and, hence, curves are nearly the same. The remaining two comparisons are for composites of the cellulose records and all 32 of the records. As can be seen, the inclusion of these additional records decreases the amplitude of the 1850-1970 decrease. We present these comparisons so that the reader has some feeling for the impact of the selection process on the shape and amplitude of the past 1850 13C/12C decline.

DECONVOLUTION PROCEDURE
The technique we use was first suggested by Siegenthaler et al. [1978]. Indeed, their preliminary calculations led these authors to foresee the main conclusions drawn here. They state, 'We find that not only the amount of biospheric CO2 is important, but, because of the time-dependent flow into the oceans, also the time of release into the atmosphere, i.e., the history of the biospheric CO2 production is essential. If a large input pulse occurred only a few decades ago, it is still influencing the present CO2 level by providing a decreasing atmospheric baseline, since CO2 is still being taken up by the ocean.' The logic behind the •3C approach is as follows. Both fossil fuel CO2 and forest-soil CO2 have •13C values averaging about -26 %o. The pre-1850 atmosphere had a value of about -6 %o. Thus addition of CO2 from either source will decrease the atmospheric •13C value. The situation is complicated by the fact that the carbon atoms from fossil fuels and forest-soil will mix with carbon atoms in the ocean and in the forest-soil reservoir. Only if the dilution caused by this mixing can properly be accounted for is the 13C/12C record for the atmosphere useful. A proponent of the large forestsoil contribution might argue that because the dilution correction is based on the same models used for the ocean uptake calculations, we are only perpetuating some basic inadequacy in these models. We will show that this is not the case. Once the dilution model has been selected (we will discuss this selection in the next section), the next step is to calculate the 13C/12C time history for atmospheric CO2, assuming that the only perturbation has come through the addition of fossil fuel CO2 to the system. As the time history of these inputs is accurately known (i.e., to -+7%) from This temporal history of forest-soil CO2 input is then obtained by iteration. An estimate of the time history is made. The 13C/12C anomaly generated by this history is calculated and compared with the residual •3C/•2C anomaly.
The history is then adjusted in such a way as to improve the match between the calculated and the observed residual.
Another run is carried out. As the shape of the observed smoothed residual is simple, this procedure proves quite effective. A good match can be achieved after only a few iterations. Once a good match is obtained, its sensitivity to  Our model is shown in Figure 4. It differs from that of Oeschger et al. [1975] in the following ways. 1. We include the oceanic photosynthesis-respiration cycle. It should be noted, however, that the inclusion of this cycle does not significantly alter the result. We include it to overcome an often invoked (but not justified) criticism of the Oeschger model. 3. The rate of vertical eddy diffusivity in the main thermocline (i.e., above the depth of the deep water source) is increased over the value adopted by Oeschger et al. [1975] so as to be consistent with the penetration of bomb-produced tritium as-determined by the GEOSECS program. The distribution of this tracer provides a better measure than does natural •4C of the extent to which substances penetrate the thermocline on a time scale of several decades. This is the most important of the modifications we have made. 4. We represent the living terrestrial biosphere (L.B. boxes in Figure 4) with three well mixed reservoirs with differing time constants. We do not include CO2-induced growth factor for these reservoirs because we are trying to account for the change in the forest and soil carbon reservoir size by the •3C/12C record deconvolution. The change in the sizes of these reservoirs is sufficiently small that there is no need to include the change in their dilution capacity with time. Were these changes to be included in the iterative procedure, they would not in any way affect the conclusions we draw. 5. We include the isotope dilution capacity of soil carbon.

We include the production of deep water. A loop is inserted that brings intermediate water to the surface in the polar region, increases its density, and sends it back to
6. Keeling et al. [1980] pointed out that the result of the type of deconvolution carried out here is dependent on the kinetic isotope fractionation factor for CO2 entering the sea. We adopt the value of 0.998 suggested by Siegenthaler and Munnich [1981] rather than that of 0.986 as derived in strongly alkaline solutions [Baertschi, 1952;Craig, 1953Craig, , 1954. However, we compare both values to show the sensitivity of isotopic fractionation factor. year period is 21 x 10 -6 atm. The model yields an increase of 23 x 10 -6 atm. As inclusion of forest-soil CO2 production would likely increase the model prediction, the short fall suggests that the model ocean is not taking up enough CO2 (or that the CO2 enhancement of forest growth is significant). An important point must be raised here. Two models of differing geometry that equally well match all the isotope distributions need not take up the same amount of CO2. The reason for this geometry dependence relates to the fact that the resistance posed by the air-sea interface is smaller for the uptake of excess CO2 than for the carbon isotope equilibra- analysis. It should be emphasized that our deconvolution should be considered as a 20-year running mean for this input. Clearly, we could create scenarios with 2-decade or less time-scale variability about the smooth scenario we have adopted that would equally well match the observations. The tree-ring curve does not provide a sufficiently detailed record to permit these second-order features to be deconvolved. It should be pointed out that the recent •3C trend for Freyer's •i•3C versus time curve is consistent with direct measurements on atmospheric CO2 samples collected over the past two decades [see Freyer and Belacy, 1983].

CALIBRATION OF THE
The atmospheric CO2 anomaly generated by the forest-soil scenario alone is shown in Figure 10. As can be seen its contribution over the last 20 years (i.e., the period of time over which a reliable atmospheric record exists) is only about 4 x 10 -6 atm. This is to be compared with a 23 x 10 -6 atm change computed by using the model for the fossil fuel contribution over this interval of time. This is one of the important consequences of the shape we obtain for the forest-soil scenario. Uptake by the ocean of CO2 released from forest and soils in earlier years has almost exactly balanced the new production from this source during the last 20 years. As can be seen in Figure 7, the same is true for •3C.
The dilution of the •3C anomaly generated by earlier inputs has almost exactly balanced the anomaly generated by the forest and soil CO2 input over the last 20 years. This As stated above, the ocean model might yield too low a ratio of excess CO2 uptake to carbon isotope dilution. The reason is that the one-dimensional model maximizes the importance of mixing relative to gas exchange resistance. While we have not yet constructed what we consider to be an adequate two-dimensional (2-D) model which takes up CO2 in regions where thermocline and deep water isopycnals outcrop, we can get a sense of the importance of this type of model modification by arbitrarily increasing the thermodynamic capacity of surface sea water for fossil fuel CO2 uptake (i.e., by reducing the Revelle factor). Such a change permits the model to take up more fossil fuel CO2 without altering its isotope dilution characteristics. In Figure 13  Finally, we compare these CO2-versus-time curves with other sources of I information (see Figure 14). During the early 1880s, the French made a series of atmospheric CO2 content measurements on an expedition through the Atlantic Ocean [Muntz and Aubin, 1886]. Unfortunately, there is no way to evaluate their absolute accuracy. Neftel et al. [1982] report a value of 271 _+ 9 ppm for the CO2 content of the atmosphere about 600 years ago, based on CO2 to air ratios in gases released from the bubbles in Greenland ice cores.
Finally, Chen and Millero [1979] estimate from measurements on 'old' ocean water that the preanthropogenic CO2 content was 275 _+ 20 ppm. As pointed out by Shiller [1981], the approach taken by these authors is subj[ct to many serious pitfalls. Thus, while all the methods give consistent values, none is reliable enough to permit a firm answer to be given regarding the magnitude of the forest-soil contribution to the build-up of atmospheric CO2. CONCLUSIONS Unfortunately, the data in hand still are not adequate to settle the controversy between terrestrial biologists who estimate the forest-soil CO2 contribution from land use and carboil inventory response data and geochemists who model the ocean uptake of excess CO2. •3C/•2C results which many hoped would serve to resolve the dilemma are still not adequate for the task. If future work bears out Freyer's composite •3C/•2C curve, then the terrestrial biologists will have been proven correct in their estimate that the total amount of excess CO2 from forests and soils has over the last 200 years exceeded that from fossil fuels. HoweveL they will have been shown to be incorrect i n their contention that the forest-soil CO2 contribution has had a strong influence on the trend in atmospheric CO2 content from 1958 to present.
Aspects of the problem still needing investigation are as follows: 1. Further work must be done to identify and eliminate sources of 'noise' in the tree ring •3C/•2C record. This is still the most powerful approach to the problem. The recent work of Farquhar et al. [1982] and Francey and Farquhar [1982] toward developing a simple theory to explain carbon isotope fractionation by plants is an important step toward understanding the meaning of the •3C/•2C record in tree rings.
2. The effect of outcropping isopycnal horizons on the ability of the ocean to take up excess CO2 (while not changing its carbon isotope dilution response) must be investigated. If this is to be done, 2-D ocean models must replace the current 1-D models.
3. CO2 partial pressure estimates from ice cores and from old photographs of solar spectra must be fully exploited.
4. The concept that land use impacts on the forest and soil carbon inventory followed population growth (at least until the early 1950s) must be reexamined. It is possible that the great expansion of grazing land which accompanied colonization had a greater impact than the expansion of human population.
5. The possibility that the biomass in the world's unimpacted forests and soils has shown a recent increase due to anthropogenic CO2, nitrogen, and phosphorus fertilization must be examined. The land-use approach disregards this countering influence. While being cut back around the 'edges,' the forests may be becoming more lush in the