Search

Since model bias and associated initialization shock are serious shortcomings that reduce prediction skills in state‐of‐the‐art decadal climate prediction efforts, we pursue a complementary machine‐learning‐based approach to climate prediction. The example problem setting we consider consists of predicting natural variability of the North Atlantic sea surface temperature on the interannual timescale in the pre‐industrial control simulation of the Community Earth System Model. While previous works have considered the use of recurrent networks such as convolutional LSTMs and reservoir computing networks in this and other similar problem settings, we currently focus on the use of feedforward convolutional networks. In particular, we find that a feedforward convolutional network with a Densenet architecture is able to outperform a convolutional LSTM in terms of predictive skill. Next, we go on to consider a probabilistic formulation of the same network based on Stein variational gradient descent and find that in addition to providing useful measures of predictive uncertainty, the probabilistic (Bayesian) version improves on its deterministic counterpart in terms of predictive skill. Finally, we characterize the reliability of the ensemble of machine learning models obtained in the probabilistic setting by using analysis tools developed in the context of ensemble numerical weather prediction. Plain Language Summary: Businesses and government agencies rely heavily on numerical predictions of climate variables such as temperature and precipitation for a wide variety of purposes ranging from integrated assessment to developing mitigation strategies to developing resilience and adaptation strategies. Developing interannual to decadal predictions using comprehensive and complex climate and earth system models, however, are computationally intensive. As such, computationally efficient and accurate surrogates of comprehensive earth system models is highly desired. Data‐driven models using advanced deep learning algorithms are promising for this purpose. This paper first considers a recently proposed convolutional network architecture to develop such a surrogate and then integrates Bayesian inference to this architecture to further assess predictive uncertainty. We show that the resulting Bayesian deep learning model not only improves prediction accuracy but also quantifies the uncertainty arising from the data and model. Key Points: Model bias and associated initialization shock are serious shortcomings that reduce prediction skill in state‐of‐the‐art decadal climate prediction effortsA complementary machine‐learning‐based approach to climate prediction is considered. Both deterministic and probabilistic machine learning approaches are examinedIn addition to providing useful measures of predictive uncertainty, Bayesian versions of deep learning models outperform their deterministic counterparts in terms of predictive skill [ABSTRACT FROM AUTHOR]

We demonstrate the use of a probabilistic machine-learning technique to develop stochastic parameterizations of atmospheric column physics. After suitable preprocessing of NASA's Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA2) data to minimize the effects of high-frequency, high-wavenumber component of MERRA2 estimate of vertical velocity, we use generative adversarial networks to learn the probability distribution of vertical profiles of diabatic sources conditioned on vertical profiles of temperature and humidity. This may be viewed as an improvement over previous similar but deterministic approaches that seek to alleviate both, shortcomings of human-designed physics parameterizations, and the computational demand of the "physics" step in climate models. [ABSTRACT FROM AUTHOR]

Reduced‐order dynamical models play a central role in developing our understanding of predictability of climate irrespective of whether we are dealing with the actual climate system or surrogate climate models. In this context, the linear inverse modeling (LIM) approach, by capturing a few essential interactions between dynamical components of the full system, has proven valuable in providing insights into predictability of the full system. We demonstrate that reservoir computing (RC), a form of learning suitable for systems with chaotic dynamics, provides an alternative nonlinear approach that improves on the predictive skill of the LIM approach. We do this in the example setting of predicting sea surface temperature in the North Atlantic in the preindustrial control simulation of a popular earth system model, the Community Earth System Model so that we can compare the performance of the new RC‐based approach with the traditional LIM approach both when learning data are plentiful and when such data are more limited. The improved predictive skill of the RC approach over a wide range of conditions—larger number of retained EOF coefficients, extending well into the limited data regime, etc.—suggests that this machine‐learning technique may have a use in climate predictability studies. While the possibility of developing a climate emulator—the ability to continue the evolution of the system on the attractor long after failing to be able to track the reference trajectory—is demonstrated in the Lorenz‐63 system, it is suggested that further development of the RC approach may permit such uses of the new approach in more realistic predictability studies. Plain Language Summary: Because of the chaotic nature of the dynamics underlying many complex systems such as weather and climate, evolution of ensembles of trajectories have to be considered in order to produce future predictions of such systems. An analysis of the dynamics of such ensembles provides insights into mechanisms that make the system predictable. Because comprehensive models of such complex systems are very costly to run, reduced‐order dynamical models of them are a useful tool in conducting such ensemble‐based predictability studies. We develop such a computationally inexpensive reduced‐order model using machine learning and show that its predictive skill is comparable to those of linear inverse model (state‐of‐the‐art) when training data are plentiful, but much better when such data are more limited. Consequently, we think that this new method has wide applicability. Furthermore, given the nonlinear nature of the new method, it has the potential to provide new insights into predictability of complex systems. Key Points: A machine‐learning (ML)‐based reduced‐order modeling method is developed to study predictability of regional SST of a popular climate modelThe method extends the state‐of‐the‐art, particularly, when data are limited and when a large number of EOF coefficients are retainedThe new method has wide applicability, could act as a climate emulator with further developments, and can give new insights [ABSTRACT FROM AUTHOR]