Record-breaking hot temperatures were observed in many places around the world in 2018, causing heat-related deaths, crop failure, wildfires and infrastructural damages. In Germany, extremely hot temperatures were accompanied by extremely low precipitation, compounding the impacts. Here we investigate spring to autumn temperature and precipitation in Germany over the historical period. We show that since measurements started in 1881, Germany has never experienced as hot and dry conditions during March to November as in 2018. We analyse the rarity of the event and illustrate that estimates of return periods for such compound extreme events are extremely high but very uncertain and strongly depend on the way they are estimated. We further investigate output from an ensemble of climate model simulations (CMIP5). Most climate models represent the distributions of temperature and precipitation in Germany and their dependence relatively well. Statistical projections of the bivariate temperature-precipitation distribution suggests that a growing season such as 2018 will become less likely at warmer global mean temperatures due to slight increases in precipitation. In contrast, climate models project an increasing likelihood of a 2018-like event and much larger uncertainties both for temperature and precipitation at different warming levels. Both observation-based scaling and climate model estimates consistently project that the compound hot and dry conditions in peak summer June–August become more likely. Overall, our results highlight the challenges associated with estimating the rarity of very extreme multivariate events and illustrate how consistent future changes in multivariate extremes can be estimated from observations only.

2018年，在世界许多地方都观察到创纪录的高温，造成与热相关的死亡，农作物歉收，野火和基础设施破坏。在德国，极高的温度伴随着极低的降水，使影响更加复杂。在这里，我们调查了历史时期德国春季至秋季的温度和降水。我们表明，自1881年开始测量以来，德国从未像2018年那样经历3月至11月的炎热干燥天气。我们分析了该事件的稀有性，并说明了此类复合极端事件的回报期估算值非常高，但非常不确定，并强烈取决于估算的方式。我们将进一步研究气候模型模拟（CMIP5）的输出。大多数气候模型代表了德国的温度和降水分布及其相对较好的依赖性。双变量温度-降水分布的统计预测表明，由于降水略有增加，在全球平均温度较高的情况下，诸如2018年的生长季节将变得不太可能。相比之下，气候模型预测类似2018年事件的可能性增加，并且在不同变暖水平下温度和降水的不确定性都更大。基于观测的标度和气候模型估计均一致地预测，在6月至8月的夏季高峰期间，炎热和干燥的复合条件变得更有可能。总体，