Basic theory and fundamental understanding help us understand that global warming will continue in the 21st century. But climate model simulations help us identify how much warming we might expect in the future and how much other climate variables like sea level will change. The degree of climate sensitivity is central to the predictions of climate change. While the range of sensitivity found in climate models is similar to the values that come from other lines of evidence, 1.5-4.5 °C for a doubling of atmospheric CO2, some argue that these estimates are too high.
In recent years, a series of studies that compared the measured change in global average temperature with the increase in CO2 found lower values of climate sensitivity than the average climate model simulates; for equilibrium climate sensitivity these studies typically suggest values that are less than 2 °C. This has led many to publicly argue that projections of future climate change should be revised downward. The conclusion that the real world is much less sensitive to CO2 emissions than climate models has been directly challenged by recent studies showing that comparing the data-based methods with climate models was fraught with inadequate treatment of statistics and examination of the sensitivity of their assumptions.
The data-based methods in question use historical temperature records as an input, calculating the sensitivity of climate by comparing the difference between CO2 warming effects early and late in the records and the change in global temperature. That comparison may be biased toward too low values, because global temperature records poorly represent changes in the Arctic and record changes in sea surface temperature instead of the faster warming air above the sea level. When those errors are compensated for in scientific studies, by appropriately sampling climate models to reflect the actual temperature record, there is no statistically significant difference in the estimates of climate sensitivity that come from data and models.
Additionally, there is some evidence that the methods used to estimate climate sensitivity from data do not properly account for other external factors that influence climate change. The cooling effect from pollution in the Northern Hemisphere might have the same cooling effect, globally averaged, as a decline in solar intensity, but different outcomes in terms of global temperature. Studies that are able to distinguish these factors are rare, but when those factors are brought into account, the average climate sensitivity from data-based methods is well inside the IPCC range.
It has proven difficult for scientists to narrow the wide range of values for climate sensitivity because of limits in the records of past changes and the relative influence of different climate agents. Different lines of evidence (e.g., basic physics, comparisons to past ice ages, and the temperature response to human influence) strongly refute sensitivities much higher or lower than the IPCC. However, viewing only one type of evidence in isolation can lead to logical errors when excluded lines of evidence refute very high and very low values.
To convincingly refine our estimates of climate sensitivity will likely require both breakthroughs in climate physics (particularly, a better understanding of how cloud cover will act to accentuate or reduce temperature trends) and enough time to pass that we can accumulate more high-quality measurements of the real-world response to today’s climate drivers. Both outcomes likely lie a couple decades hence[3], until then decisions will have to be made with some ambiguity in the potential climate response.