By J.B. Marston, Laboratory of Atomic and Solid State Physics, Cornell University;
M. Oppenheimer, R. M. Fujita, S. R. Gaffin, Environmental Defense Fund
Scientific Correspondence reprinted from Nature, Vol. 349.
February 14, 1991
Dear Sir:
Kuo et al. have shown
[1] that the monthly concentration of atmospheric carbon dioxide at Mauna Loa, Hawaii exhibits statistically significant coherences over a range of frequencies with monthly surface air temperatures averaged over the entire globe. The CO2 record lags behind the temperature record; this lag is consistent with the hypothesis that temperature fluctuations or associated meteorological changes [2] cause the short term CO2 anomalies rather than vice versa. In any event, fluctuations in CO2 concentration at frequencies > 0.2 per year are too small to cause the observed temperature variations by the greenhouse effect. We have analyzed air and sea surface temperatures averaged over different regions of the Earth, and find that air temperatures averaged over equatorial regions and sea temperatures averaged over the equatorial Pacific are more highly correlated with the CO2 anomalies than air temperatures averaged over non- equatorial northern (90° to 23.6° N) or southern (90° to 23.6° S) regions. This result suggests that the sources and sinks of atmospheric CO2 in question lie in the equatorial regions.We have used the Goddard Institute for Space Studies (GISS) monthly air temperature record from land stations,
[3] monthly sea surface temperature (SST) data from the Comprehensive Ocean Atmospheric Data Set (COADS) [4] and the monthly atmospheric CO2 record at Mauna Loa to search for coherences between the temperature and the month to month change in the CO2 concentration. The seasonal cycle in both the CO2 and temperature time series is removed by calculating the departures from the mean values for each month of the year. Spectral analysis allows us to ascertain directly the degree of coherency between the monthly temperatures and the monthly rate of change in CO2 at different frequencies. We use a gaussian spectral window of half- width o = 0.07 per year and apply a smooth cutoff to each end of the time series to reduce leakage.Insert graph environmentaldefense.org/pubs/reports/co2tempgraph.gif
Part a of the figure shows the coherency between the monthly rate of change of CO2 and surface air temperatures averaged over the entire globe, over non- equatorial northern regions and over non-equatorial southern regions during March 1958 to December 1987. The probability that the observed coherences between the global air temperature and the month to month change in CO2 in the frequency range 0.0 cycles per year < v < 0.55 per year occurred by chance is less than 5%. The rise in the coherency at frequencies < 0.15 per year is due to long-term increasing trends in both series. As the non-equatorial northern and southern regions show much weaker coherency for v > 0.1 cycles per year, we conclude that most of the short-term correlations present in the global temperature record originate in the equatorial region. Indeed, equatorial air temperatures from 23.6° S to 23.6° N are coherent with CO2, over a broader range of frequencies than temperatures averaged over the entire globe (b in the figure). Coherences in the equatorial record at frequencies higher than roughly 0.5 cycles per year are apparently washed out in the global average. Sea surface temperatures averaged over the equatorial Pacific region (20° S to 20° N and 80° W to 180°) show coherences comparable to those in the equatorial air temperature record at frequencies greater than about 0.25 per year. Both the equatorial air and the sea surface temperature show three peaks in the coherency at frequencies of approximately 0.3,,0.5 and 0.8 cycles per year, but we do not know if they represent physically distinct processes.
Phase analysis shows that the SST fluctuations lead the CO2 anomalies by about 4 months but equatorial air temperature leads CO2 by only 1 month. This difference is not surprising as the air temperature is strongly correlated with the SST but lags behind by about 3 months. The relative time lags are consistent with the hypothesis that fluctuations in SST set off changes in the equatorial air temperature and the CO2 concentration. If SST variations directly affect equatorial sources and sinks then the lag is consistent with the atmospheric transport time for a parcel of CO2- enriched or CO2- depleted air to reach Mauna Loa.
The flux of CO2 between the atmosphere and the equatorial regions can be affected by fluctuations in the temperature through changes in CO2 solubility, upwelling of CO2-enriched deep water, and availability of nutrients for the organic carbon pump. Associated changes in wind and precipitation also affect terrestrial biota. The coherence between temperature and CO2 is significant over a range of frequencies in addition to those that characterize ENSO. [2] It is important to consider whether these mechanisms could lead to a significant positive temperature- CO2 feedback on the longer timescales for which global warming is projected to occur.
FOOTNOTES
1. Kuo, C., Lindberg, C. & Thomson, D. Nature 343: 709- 713 (1990). Back to Text.
2 . Keeling, C. D. et. al. Geophys. Monog.. 55:165-236 (1989). Back to Text.
3 . Hansen, J. & Lebedeff, S. Geophys Res. Lett. 15: 323- 326 (1988). Back to Text.
4. Oort, A.H., Pan, R.W., Reynolds, R.W. & Ropelewski, C.F. Climate Dynamics 2: 29- 38 (1987) . Back to Text.
5. Bloomfield, P. Fourier Analysis of TIme Series: An Introduction (Wiley, New York,1976). Back to Text.