Causal discovery of Atlantic-Pacific interactions in observations and CMIP6 models

Abstract

In a world of rapid climate change, a deeper comprehension of the Earth’s climate system is central to accurate climate projections. In this regard, understanding the dynamics governing the climate system is important. The changing climate processes are influenced by a variety of factors including both natural and anthropogenic forcings, which modulate the interactions between major modes of climate variability. These interactions, particularly the teleconnections between the Atlantic and Pacific oceans, have a profound impact on global and regional climate patterns, necessitating a detailed exploration to grasp the complex networks of interrelated impacts. In this thesis, causal discovery approaches are applied to unravel the causal relationships for these interactions, aiming to enhance the understanding of the processes governing the climate system. The first part of this Ph.D. thesis delves into this complex system by applying an algorithm for causal discovery to analyze observational and reanalysis datasets, in addition to large ensemble simulations from a collection of models participating in the sixth phase of the Coupled Model Intercomparison Project (CMIP). Dependent on the phases of the Pacific Decadal Variability (PDV) and the Atlantic Multidecadal Variability (AMV), different regimes with characteristic causal relationships (fingerprints) are identified in observations (and reanalyses) as well as in CMIP6. A regime-oriented causal model evaluation is then performed to assess the capability of CMIP6 models in representing observed changing interactions between PDV, AMV, and their extratropical teleconnections. Causal networks from observations show both opposite-sign and same-sign responses between AMV and PDV under specific conditions. Historical CMIP6 simulations exhibit varying skill in simulating the observed causal patterns but overall perform better when PDV and AMV are out of phase. Additionally, the two largest ensembles, (in terms of number of realizations) were found to contain realizations with most similar causal fingerprints to observations. For most regimes, these same models also showed higher network similarity when compared to each other. In the second part of this thesis, the focus is on examining the tropical and extra-tropical routes connecting Pacific and Atlantic modes of variability on seasonal to interannual timescales. Following up on recent studies, this analysis characterizes two distinctive phases: the Pacificdriven regime (1950-1983) and theAtlantic-driven regime (1985-2014), spotlighting the varying role of El Niño-Southern Oscillation (ENSO) in shaping sea surface temperature variability in the tropical Atlantic. Guided by the results of the first study, the use of large ensemble simulations in this second study intends to separate the contributions of external forcings from natural internal variability. A comparative analysis examines results from observations (and reanalysis) in contrast to those from Pacific pacemaker simulations, unveiling effects of anthropogenic external forcing, especially in the most recent decades. Specifically, the 1985-2014 results suggest that human-induced anomalous tropical north Atlantic warming greatly contributed to La Niña-like cooling over the tropical Pacific through the strengthening of the Pacific Walker circulation. On the other hand, the causal analysis of the pre-industrial control run emphasizes the importance of natural internal variability on decadal timescales in modulating the interplay between interannual climate variability modes over the two basins. Generally, the results presented in this thesis demonstrate the large potential of causal discovery for process-oriented model evaluation that can substantially enhance our understanding of climate variability and provide robust diagnostics for refining climate models. Furthermore, this thesis underscores the role of the intricate interplay between natural variability and external forcings in shaping climate patterns, and advocates for further research to precisely attribute the observed changes in the climate system. The insights gained are hence significant for formulating more accurate and informed climate projections as well as adaptation and mitigation strategies.