The CO2 climate response problem a statistical approach

Abstract

The anthropogenic increase of the atmospheric carbon dioxide (CO2) concentration leads to a global warming of the atmospheric surface layer, whereas the stratosphere is cooled. This “greenhouse” effect postulated by a number of climate models (on a physical basis) can be conditionally verified by statistical multiple regression techniques. In this study the following climatic time series are used (all data yearly averages): northern hemisphere mean temperatures near surface 1781–1980 (alternatively since 1851 or 1881) and corresponding stratospheric data 1958–1983, sea surface temperatures 1856–1980, northern hemisphere or global average, alternatively, and the global mean sea level fluctuations 1881–1980. In order to account for an appropriate part of explained variance, volcanic and solar forcing parameters are implied and the data are low-pass filtered suppressing variations of the period rangeT < 10 years. Based on the recently assessed preindustrial CO2 concentration of c. 280 ppm and the Mauna Loa value of c. 344 ppm in 1984 this “industrial” CO2 increase reveals a northern hemisphere temperature increase near surface of c. (.7±.1) K (average and standard deviation of all statistical regression runs), statistically significant at the 95% level. A CO2 doubling (300 to 600 ppm) leads to a statistically derived signal of (3.1±.6) K, satisfactorily congruent with the results of most of the (deterministic) climate models: c. (3±1.5)K. A stratospheric cooling trend in recent time may be existent but is highly non-significant. Similarly, the SST data do not allow to evaluate a significant CO2 signal to noise ratio. In contrast to that the observed long-term global mean sea level increase (9.3 cm) can be predominantly attributed to the CO2 effect (99.9% level).