The discovery of cause-effect relationships helps to understand the natural or physical mechanism [1].Causation inference is a key issue in many disciplines and has a long study history,especially in statistics,social,and biomedical sciences [2].In Earth System Sciences,the cause-effect relationship also plays a fundamental role and has drawn increasing interests.However,for spatially large-scale research,it is not feasible to design and conduct con-trolled experiments to reveal the cause-effect relationships.There-fore,causation inference from time series data has been frequently employed,under the assumption that the cause precedes the effect[3].While the temporal inference works effectively to identify most causation between variables,limitations remain.If the time series is not long enough to catch significant changes of causes and effects,some important cause-effect relationships may be neglected.This limitation is highlighted in Earth System Sciences,as the evolution of global changes may take an extreme long period to present discernible variations.For instance,the annually mean temperature in one area demonstrates very limited variations in decades,which is already a long period for Earth observation.On one hand,the causation between temperature and plant growth in this area can hardly be identified using temporal causation mod-els,which mainly detect causation between two variables by examining the successive (or simultaneous) variations of one vari-able induced by the variation of the other.On the other hand,the causal influence of temperature on plant growth has been well accepted [4,5].So clearly,the temporal causality models are not a panacea to all causation inference scenarios.