A novel big-data perspective on earth system evolution

The many components of the Earth System are linked through complex feedbacks that can be revealed in large compilations of diverse geochemical proxy data for paleoclimate and ocean conditions. In those archives, for example, temporal and spatial trends can be visualized as topologies (‘landscapes’) defined by areas of high-density data, corresponding to steady states (= stability basins) maintained by negative feedbacks. The boundaries between those stability basins, representing low-density data areas (potential tipping points), are crossed when external drivers are involved, moving the system to a new steady state. These external drivers are often associated with positive feedbacks. As a proof of concept, we produced a stability ‘landscape’ using an extensive set of published carbon (δ13C) and oxygen (δ18O) isotope data from sedimentary carbonates spanning the last 2.5 billion years. The superimposed C-O isotopic pathways show a preference for particular regions of the ‘landscape’ at different times in Earth history. Major excursions reflect positive loops often set into motion by external inputs (drivers) that can overwhelm the system, such as major volcanic and tectonic events and human-induced climate effects. Our approach can be applied to other proxy datasets, and multivariate statistical treatments, including machine-learning approaches, can potentially yield a robust, high-resolution ‘landscape’ of the Earth System through time.