Continuous structural parameterization: a proposed method for representing different model parameterizations within one structure demonstrated for atmospheric convection
Journal of Advances in Modeling Earth Systems American Geophysical Union 12:8 (2020) e2020MS002085
Abstract:
Continuous structural parameterization (CSP) is a proposed method for approximating different numerical model parameterizations of the same process as functions of the same grid‐scale variables. This allows systematic comparison of parameterizations with each other and observations or resolved simulations of the same process. Using the example of two convection schemes running in the Met Office Unified Model (UM), we show that a CSP is able to capture concisely the broad behavior of the two schemes, and differences between the parameterizations and resolved convection simulated by a high resolution simulation. When the original convection schemes are replaced with their CSP emulators within the UM, basic features of the original model climate and some features of climate change are reproduced, demonstrating that CSP can capture much of the important behavior of the schemes. Our results open the possibility that future work will estimate uncertainty in model projections of climate change from estimates of uncertainty in simulation of the relevant physical processes.The value of initialisation on decadal timescales: state dependent predictability in the CESM Decadal Prediction Large Ensemble
Journal of Climate American Meteorological Society 33:17 (2020) 7353-7370
Abstract:
Information in decadal climate prediction arises from a well initialised ocean state and from the predicted response to an external forcing. The length of time over which the initial conditions benefit the decadal forecast depends on the start date of the forecast. We characterise this state-dependent predictability for decadal forecasts of upper ocean heat content in the Community Earth System Model. We find regionally dependent initial condition predictability, with extended predictability generally observed in the extra-tropics. We also detect state-dependent predictability, with the year of loss of information from the initialisation varying between start dates. The decadal forecasts in the North Atlantic show substantial information from the initial conditions beyond the ten-year forecast window, and a high degree of state-dependent predictability. We find some evidence for state dependent predictability in the ensemble spread in this region, similar to that seen in weather and subseasonal-to-seasonal forecasts. For some start dates, an increase of information with lead time is observed, for which the initialised forecasts predict a growing phase of the Atlantic Multidecadal Oscillation. Finally we consider the information in the forecast from the initial conditions relative to the forced response, and quantify the crossover timescale after which the forcing provides more information. We demonstrate that the climate change signal projects onto different patterns than the signal from the initial conditions. This means that even after the crossover timescale has been reached in a basin-averaged sense, the benefits of initialisation can be felt locally on longer timescales.Euro-Atlantic weather Regimes in the PRIMAVERA coupled climate simulations: impact of resolution and mean state biases on model performance
Climate Dynamics Springer Science and Business Media LLC 54:11-12 (2020) 5031-5048
Machine Learning for Stochastic Parameterization: Generative Adversarial Networks in the Lorenz’96 Model
Journal of Advances in Modeling Earth Systems American Geophysical Union 12:3 (2020) e2019MS001896
Abstract:
Stochastic parameterizations account for uncertainty in the representation of unresolved subgrid processes by sampling from the distribution of possible subgrid forcings. Some existing stochastic parameterizations utilize data‐driven approaches to characterize uncertainty, but these approaches require significant structural assumptions that can limit their scalability. Machine learning models, including neural networks, are able to represent a wide range of distributions and build optimized mappings between a large number of inputs and subgrid forcings. Recent research on machine learning parameterizations has focused only on deterministic parameterizations. In this study, we develop a stochastic parameterization using the generative adversarial network (GAN) machine learning framework. The GAN stochastic parameterization is trained and evaluated on output from the Lorenz '96 model, which is a common baseline model for evaluating both parameterization and data assimilation techniques. We evaluate different ways of characterizing the input noise for the model and perform model runs with the GAN parameterization at weather and climate time scales. Some of the GAN configurations perform better than a baseline bespoke parameterization at both time scales, and the networks closely reproduce the spatiotemporal correlations and regimes of the Lorenz '96 system. We also find that, in general, those models which produce skillful forecasts are also associated with the best climate simulations.Constraining stochastic parametrisation schemes using high-resolution simulations
Quarterly Journal of the Royal Meteorological Society Wiley (2019)