The most direct method for measuring atmospheric carbon dioxide concentrations for periods before direct sampling is to measure bubbles of air (fluid or gas inclusions) trapped in the Antarctic or Greenland ice sheets. The most widely accepted of such studies come from a variety of Antarctic cores and indicate that atmospheric CO2 levels were about 260–280 ppmv immediately before industrial emissions began and did not vary much from this level during the preceding 10,000 years (10 ka). In 1832 Antarctic ice core levels were 284 ppmv.
One study disputed the claim of stable CO2 levels during the present interglacial of the last 10 ka. Based on an analysis of fossil leaves, Wagner et al. argued that CO2 levels during the period 7–10 ka were significantly higher (~300 ppm) and contained substantial variations that may be correlated to climate variations. Others have disputed such claims, suggesting they are more likely to reflect calibration problems than actual changes in CO2. Relevant to this dispute is the observation that Greenland ice cores often report higher and more variable CO2 values than similar measurements in Antarctica. However, the groups responsible for such measurements (e.g. H. J Smith et al.) believe the variations in Greenland cores result from in situ decomposition of calcium carbonate dust found in the ice. When dust levels in Greenland cores are low, as they nearly always are in Antarctic cores, the researchers report good agreement between Antarctic and Greenland CO2 measurements.
The longest ice core record comes from East Antarctica, where ice has been sampled to an age of 800 ka. During this time, the atmospheric carbon dioxide concentration has varied between 180–210 ppm during ice ages, increasing to 280–300 ppm during warmer interglacials. The beginning of human agriculture during the current Holocene epoch may have been strongly connected to the atmospheric CO2 increase after the last ice age ended, a fertilization effect raising plant biomass growth and reducing stomatal conductance requirements for CO2 intake, consequently reducing transpiration water losses and increasing water usage efficiency.
On long timescales, atmospheric CO2 content is determined by the balance among geochemical processes including organic carbon burial in sediments, silicate rock weathering, and volcanism. The net effect of slight imbalances in the carbon cycle over tens to hundreds of millions of years has been to reduce atmospheric CO2. On a timescale of billions of years, such downward trend appears bound to continue indefinitely as occasional massive historical releases of buried carbon due to volcanism will become less frequent (as earth mantle cooling and progressive exhaustion of internal radioactive heat proceeds further). The rates of these processes are extremely slow; hence they are of no relevance to the atmospheric CO2 concentration over the next hundreds, thousands, or millions of years.
Various proxy measurements have been used to attempt to determine atmospheric carbon dioxide levels millions of years in the past. These include boron and carbon isotope ratios in certain types of marine sediments, and the number of stomata observed on fossil plant leaves. While these measurements give much less precise estimates of carbon dioxide concentration than ice cores, there is evidence for very high CO2 volume concentrations between 200 and 150 Ma of over 3,000 ppm and between 600 and 400 Ma of over 6,000 ppm. In more recent times, atmospheric CO2 concentration continued to fall after about 60 Ma. About 34 Ma, the time of the Eocene-Oligocene extinction event and when the Antarctic ice sheet started to take its current form, CO2 is found to have been about 760 ppm, and there is geochemical evidence that volume concentrations were less than 300 ppm by about 20 Ma. Carbon dioxide decrease, with a tipping point of 600 ppm, was the primary agent forcing Antarctic glaciation. Low CO2 concentrations may have been the stimulus that favored the evolution of C4 plants, which increased greatly in abundance between 7 and 5 Ma.
Assuming a future absence of human impact influencing releasing of sequestered carbon, the long term natural trend is for the plant life on land to die off altogether, as most of the remaining carbon in the atmosphere will become sequestered underground on a billion-years timescale, as natural releases of CO2 by radioactivity-driven tectonic activity will continue to slow down. Some microbes are capable of photosynthesis at concentrations of CO2 of a few parts per million. Last life forms would probably disappear only because of rising temperatures and loss of the atmosphere when the sun becomes a red giant some four-billion years from now. The loss of plant life will also result in the eventual loss of oxygen (see also Future of the Earth).
Read more about this topic: Carbon Dioxide In Earth's Atmosphere