Journal Article

Improving the Quasi‐Biennial Oscillation via a Surrogate‐Accelerated Multi‐Objective Optimization

Debusschere, Bert J., Hannah, Walter, Eldred, Michael S., Chen, Chih‐Chieh, Damiano, Luis, Benedict, James J., Sargsyan, Khachik

2025

Overview

Accurate simulation of the quasi‐biennial oscillation (QBO) is challenging due to uncertainties in representing convectively generated gravity waves. We develop an end‐to‐end uncertainty quantification workflow that calibrates these gravity wave processes in E3SM for a realistic QBO. Central to our approach is a domain knowledge‐informed, compressed representation of high‐dimensional spatio‐temporal wind fields. By employing a parsimonious statistical model that learns the fundamental frequency from complex observations, we extract interpretable and physically meaningful quantities capturing key attributes. Building on this, we train a probabilistic surrogate model that approximates the fundamental characteristics of the QBO as functions of critical physics parameters governing gravity wave generation. Leveraging the Karhunen–Loève decomposition, our surrogate efficiently represents these characteristics as a set of orthogonal features, capturing cross‐correlations among multiple physics quantities evaluated at different pressure levels and enabling rapid surrogate‐based inference at a fraction of the computational cost of full‐scale simulations. Finally, we analyze the inverse problem using a multi‐objective approach. Our study reveals a tension between amplitude and period that constrains the QBO representation, precluding a single optimal solution. To navigate this, we quantify the bi‐criteria trade‐off and generate a set of Pareto optimal parameter values that balance the conflicting objectives. This integrated workflow improves the fidelity of QBO simulations and offers a versatile template for uncertainty quantification in complex geophysical models. Plain Language Summary Simulating the quasi‐biennial oscillation (QBO), a regular pattern of alternating winds high in the atmosphere, remains a major challenge for climate models. We developed an end‐to‐end workflow to calibrate gravity wave processes in the Energy Exascale Earth System Model, leading to more realistic simulations. We began by compressing complex spatio‐temporal data into a few key, physically meaningful quantities, such as the oscillation's amplitude and period. This data reduction allowed us to isolate the QBO signal from noise and other atmospheric phenomena. Next, we built a fast statistical model that predicts QBO behavior based on critical physics parameters. This surrogate efficiently captures relationships among various atmospheric features, reducing the need for computationally expensive full‐scale simulations. Our analysis revealed a trade‐off between QBO amplitude and period, meaning that improving one aspect often worsened the other. Rather than finding a single perfect solution, we identified a range of balanced settings that offer the best compromise. This integrated approach not only leads to more realistic QBO simulation but also provides a practical framework for tuning other complex atmospheric phenomena. Key Points We developed an end‐to‐end workflow that calibrates gravity wave generation in E3SMv3, improving quasi‐biennial oscillation (QBO) realism The fundamental frequency model compressed wind field data into physically interpretable quantities, isolated the QBO signal, and reduced dimensionality while retaining key QBO variability Our workflow reveals no single optimal configuration for QBO realism, but a frontier of best‐compromise solutions

• Publisher: John Wiley & Sons, Inc,American Geophysical Union (AGU)
• Subject: Automation / Calibration / dimension reduction / fundamental frequency model / General circulation models / Gravity waves / model calibration / Optimization / Pareto frontier / Physics / probabilistic surrogate / Quasi-biennial oscillation / Sensitivity analysis / Statistical models / Stratosphere / Uncertainty / Wave generation / Wind fields / Wind shear