David Victor and Charles Kennel challenge the practice of using global mean temperature as the main measure of danger from climate change (Nature 514, 30–31; 2014). On the basis of 40 years of science and policy research, there are good reasons why this temperature is the favoured indicator.

It can be related through climate models to the regional impacts and risks that drive public concern (see go.nature.com/5chktj). It is indeed “related only probabilistically to emissions”, but the authors' best indicator — carbon dioxide concentration — is related only probabilistically to impacts and risks, except in the case of ocean acidification. As for ocean heat content, its trend experiences interruptions much like the global mean temperature, and bears no direct relationship to most impacts and risks.

Compared with other proposals, global mean temperature is more closely related to outcomes for people and ecosystems. Without such a goal, we shall never know how much reduction in emissions is sufficient.