Abstract:Modern climate projections often suffer from inadequate spatial and temporal resolution due to computational limitations, resulting in inaccurate representations of sub-grid processes. A promising technique to address this is the Multiscale Modeling Framework (MMF), which embeds a kilometer-resolution cloud-resolving model within each atmospheric column of a host climate model to replace traditional convection and cloud parameterizations. Machine learning (ML) offers a unique opportunity to make MMF more accessible by emulating the embedded cloud-resolving model and reducing its substantial computational cost. Although many studies have demonstrated proof-of-concept success of achieving stable hybrid simulations, it remains a challenge to achieve near operational-level success with real geography and comprehensive variable emulation that includes, for example, explicit cloud condensate coupling. In this study, we present a stable hybrid model capable of integrating for at least 5 years with near operational-level complexity, including real geography, seasonality, explicit cloud condensate predictions, and land coupling. Our model demonstrates skillful online performance in metrics such as 5-year zonal mean biases compared to previous MMF emulation studies. The monthly error against reference MMF simulations with the same initial condition approaches the fundamental predictability limit. Key factors contributing to our online performance include an expressive U-Net architecture, additional input features that include large-scale forcings and convection memory, and physical thermodynamic constraints for microphysics. With microphysical constraints mitigating unrealistic cloud formation, our work is the first to demonstrate realistic multi-year cloud condensate climatology under the MMF framework. Our work showcases ML parameterization's potential for operational-level climate simulations.

Computationally-Efficient Climate Predictions using Multi-Fidelity Surrogate Modelling

Adaptive Multifidelity Data Assimilation for Nonlinear Subsurface Flow Problems

Graph Convolutional Neural Networks as Surrogate Models for Climate Simulation

A Reliable Climate Model Based on Gaussian Process

Improving Met Office Weather and Climate Forecasts with Bespoke Multigrid Solvers

Efficient surrogate modeling methods for large-scale Earth system models based on machine learning techniques

Climate Change Modelling at Reduced Float Precision with Stochastic Rounding

Multi-fidelity Hierarchical Neural Processes

Multi-fidelity climate model parameterization for better generalization and extrapolation

Probabilistic Emulation of a Global Climate Model with Spherical DYffusion

Machine learning emulation of precipitation from km-scale regional climate simulations using a diffusion model

Compression and Conditional Emulation of Climate Model Output

Prediction and Computer Model Calibration Using Outputs From Multi-fidelity Simulators

Advances in Land Surface Model-based Forecasting: A comparative study of LSTM, Gradient Boosting, and Feedforward Neural Network Models as prognostic state emulators

Machine learning emulation of a local-scale UK climate model

Spatiotemporally Coherent Probabilistic Generation of Weather from Climate

Stable Machine-Learning Parameterization of Subgrid Processes with Real Geography and Full-physics Emulation

Multi-fidelity Gaussian process surrogate modeling for regression problems in physics

Upskilling low‐fidelity hydrodynamic models of flood inundation through spatial analysis and Gaussian Process learning

Predicting global patterns of long-term climate change from short-term simulations using machine learning

A cheap data assimilation approach for expensive numerical simulations