The complete or partial collapse of the forests of Amazonia is consistently named as one of the top ten possible tipping points of Planet Earth in a changing climate. However, apart from a few observational studies that showed increased mortality after the severe droughts of 2005 and 2010, the evidence for such collapse depends primarily on modelling. Such studies are notoriously deficient at predicting the rainfall in the Amazon basin and how the vegetation interacts with the rainfall is poorly represented. Here, we use long-term surface-based observations of the air temperature and rainfall in Amazonia to provide a constraint on the modelled sensitivity of temperature to changes in precipitation. This emergent constraint also allows us to significantly constrain the likelihood of a forest collapse or dieback. We conclude that Amazon dieback under IPCC scenario RCP8.5 (crossing the tipping point) is not likely to occur in the twenty-first century.

在不断变化的气候中，亚马孙森林的全部或部分倒塌一直被誉为地球上十大可能的引爆点之一。但是，除了一些观察性研究显示2005年和2010年严重干旱后死亡率增加外，这种崩溃的证据主要取决于模型。众所周知，此类研究不足以预测亚马逊河流域的降雨，而且植被与降雨之间的相互作用方式也很不完整。在这里，我们对亚马逊地区的气温和降雨量进行了长期的基于地面的观测，从而为温度对降水变化的敏感性建模提供了约束。这种紧急约束条件还使我们能够极大地限制森林倒塌或倒退的可能性。