Prediction of climate tipping is challenging due to the lack of recent observation of actual climate tipping. Despite many previous efforts to accurately predict the existence and timing of climate tippings under specific climate scenarios, the predictability of climate tipping, the necessary conditions under which climate tipping can be predicted, has yet to be explored. In this study, the predictability of climate tipping is analyzed by Observation System Simulation Experiment (OSSE), in which the value of observation for prediction is assessed through the idealized experiment of data assimilation, using a simplified dynamic vegetation model and an Atlantic Meridional Overturning Circulation (AMOC) two box model. We find that the ratio of internal variability to observation error, or signal-to-noise ratio, should be large enough to accurately predict climate tippings. When observation can accurately resolve the internal variability of the system, assimilating these observations into process-based models can effectively improve the skill of predicting climate tippings. Our quantitative estimation of required observation accuracy to predict climate tipping implies that the existing observation network is not always sufficient to accurately project climate tipping.

翻译：由于缺乏近期实际气候临界点的观测数据，气候临界点的预测面临挑战。尽管先前已有诸多研究致力于在特定气候情景下准确预测气候临界点的存在与发生时间，但关于气候临界点可预测性——即能够预测气候临界点的必要条件——的探讨尚不充分。本研究通过观测系统模拟实验（OSSE）分析气候临界点的可预测性，该实验利用简化的动态植被模型与大西洋经向翻转环流（AMOC）双箱模型，通过数据同化的理想化实验评估观测对预测的价值。研究发现，系统内部变率与观测误差的比值（即信噪比）必须足够大才能准确预测气候临界点。当观测能够精确解析系统内部变率时，将这些观测数据同化到基于过程的模型中，可有效提升气候临界点的预测能力。我们对预测气候临界点所需观测精度的定量评估表明，现有观测网络尚不足以始终满足准确预测气候临界点的需求。