Current climate models are calibrated during their development process to match observations within reasonable uncertainty ranges. However, the warming to date due to greenhouse gas increases has been partially compensated by an uncertain amount of cooling caused by human-induced enhancement of light scattering by aerosols and by their effect on clouds; this compensation has been estimated to be from 20 to 70 percent (with 90 percent confidence) based on a range of observational and model-based studies (IPCC, 2007d). Over the 21st century, global aerosol emissions are expected to not increase further, but greenhouse gas emissions are likely to accelerate for at least the next few decades, so this compensation will become less significant. Because of the uncertain cooling by aerosols the current warming cannot be used to constrain the “climate sensitivity.” Thus, the simulated 21st-century global-average warming varies across the international suite of climate models with a range of approximately 30 percenta as is further discussed in Chapter 4.
Models provide quantitative estimates of future climate change, but with significant sources of uncertainty—lack of knowledge, or imperfect knowledge about specific quantities or the behavior of a system. These include the uncertainty in the “forcing” on the climate system from future greenhouse gas and aerosol emissions, as well as natural processes such as volcanic eruptions and solar variability, used as inputs to climate models; the uncertainty in the climate system response to this forcing; the uncertainty from natural internal variability of the climate system; the uncertainty from incomplete representations of known but complicated and small-scale processes (such as cumulus clouds) and of poorly understood processes (such as ice nucleation in clouds); and the uncertainty from “unknown unknowns” (see Chapter 6 for more information on uncertainty).
a More specifically, the interquartile range is 30 percent of the mean, where the interquartile range is a measure of statistical dispersion, and measures the difference between the 75th percentile and the 25th percentile of the data.