Quantifying and characterizing climate nonstationarities in response to forcing across the CESM1 Large Ensemble, Forced Ensembles, and observational products

Understanding the time evolution of climate statistics and extremes under forced and internal climate variability is important for predicting future changes in climate variability and for applications in ensemble climate data assimilation. Although past work has examined the evolution of individual climate modes in response to forcing, we do not yet have a comprehensive understanding of expected changes in second-moment statistics (covariances). This work uses output from the NCAR CESM1 Large Ensemble and Single-Forcing Ensembles to quantify and characterize aggregate changes in the spatial covariance of surface air temperature. We find that changes in covariances computed across ensemble members through time (diagnosed as a time series of Kullback-Leibler divergences of forced simulations relative to controls) reflect time series of external forcing and exceed background levels of nonstationarity diagnosed from control simulations. By computing the time evolution of Linear Inverse Models across ensemble members, we relate covariance changes to evolving dynamics of internal variability. Finally, we compare model-based results to available observation-based reanalyses and gridded instrumental products to evaluate consistency in climate variability changes among climate models and observations. These tools and analyses provide a general basis for comparison among large ensembles from different models.