Search for: All records

Abstract Two key challenges in the development of data‐driven gravity‐wave parameterizations are generalization, how to ensure that a data‐driven scheme trained on the present‐day climate will continue to work in a new climate regime, and calibration, how to account for biases in the “host” climate model. Both problems depend fundamentally on the response to out‐of‐sample inputs compared with the training dataset, and are often conflicting. The ability to generalize to new climate regimes often goes hand in hand with sensitivity to model biases. To probe these challenges, we employ a one‐dimensional (1D) quasibiennial oscillation (QBO) model with a stochastic source term that represents convectively generated gravity waves in the Tropics with randomly varying strengths and spectra. We employ an array of machine‐learning models consisting of a fully connected feed‐forward neural network, a dilated convolutional neural network, an encoder–decoder, a boosted forest, and a support‐vector regression model. Our results demonstrate that data‐driven schemes trained on “observations” can be critically sensitive to model biases in the wave sources. While able to emulate accurately the stochastic source term on which they were trained, all of our schemes fail to simulate fully the expected QBO period or amplitude, even with the slightest perturbation to the wave sources. The main takeaway is that some measures will always be required to ensure the proper response to climate change and to account for model biases. We examine one approach based on the ideas of optimal transport, where the wave sources in the model are first remapped to the observed one before applying the data‐driven scheme. This approach is agnostic to the data‐driven method and guarantees that the model adheres to the observational constraints, making sure the model yields the right results for the right reasons. 

Abstract The ensemble forecast dominates the computational cost of many data assimilation methods, especially for high‐resolution and coupled models. In situations where the cost is prohibitive, one can either use a lower‐cost model or a lower‐cost data assimilation method, or both. Ensemble optimal interpolation (EnOI) is a classical example of a lower‐cost ensemble data assimilation method that replaces the ensemble forecast with a single forecast and then constructs an ensemble about this single forecast by adding perturbations drawn from climatology. This research develops lower‐cost ensemble data assimilation methods that add perturbations to a single forecast, where the perturbations are obtained from analogs of the single model forecast. These analogs can either be found from a catalog of model states, constructed using linear combinations of model states from a catalog, or constructed using generative machine‐learning methods. Four analog ensemble data assimilation methods, including two new ones, are compared with EnOI in the context of a coupled model of intermediate complexity: Q‐GCM. Depending on the method and on the physical variable, analog methods can be up to 40% more accurate than EnOI.