The estimates of climate-change impact include large uncertainties, therefore an ensemble of projections is presented. The ensembles contain a spread of values that reflect the lack of knowledge, for instance about initial conditions, sensitivity of processes, future emissions and natural variability.

Most of the uncertainty in near-time projections refers to natural variability, which is difficult to describe in numerical models. To distinguish the climate signal from the natural variability, many model simulations are used in an ensemble to determine the climate change component. The problem of distinguishing trends from high variability is often called the signal-to-noise issue.  

For longer time scales, however, most uncertainty refers to future concentrations of greenhouse gases in the atmosphere (RCPs and SSPs), which depend on scenarios of societal evolution and implementation of mitigation measures. The scenarios determine the greenhouse gases that drive climate change in a climate model. Different climate models may respond differently to these changes, leading to an uncertainty related to the particular model used for investigating climate change.

Hence, the spread in an ensemble of models can be used both to explore climate in light of uncertainties from natural variability and serve for investigating climate model uncertainty.