Frederick Iat-Hin Tam, Fabien Augsburger, Tom Beucler
Reliably identifying and understanding temporal precursors to extreme wind gusts is crucial for early warning and mitigation. This study proposes a simple data-driven approach to extract key predictors from a dataset of historical extreme European winter windstorms and derive simple equations linking these precursors to extreme gusts over land. A major challenge is the limited training data for extreme events, increasing the risk of model overfitting. Testing various mitigation strategies, we find that combining dimensionality reduction, careful cross-validation, feature selection, and a nonlinear transformation of maximum wind gusts informed by Generalized Extreme Value distributions successfully reduces overfitting. These measures yield interpretable equations that generalize across regions while maintaining satisfactory predictive skill. The discovered equations reveal the association between a steady drying low-troposphere before landfall and wind gust intensity in Northwestern Europe.
Saranya Ganesh S, Frederick Iat-Hin Tam, Milton S. Gomez, Marie McGraw, Mark DeMaria, Kate Musgrave, Jakob Runge, Tom Beucler
Improving statistical forecasts of tropical cyclone (TC) intensity is limited by complex nonlinear interactions and difficulty in identifying relevant predictors. Conventional methods prioritize correlation or fit, often overlooking confounding variables and limiting generalizability to unseen TCs. To address this, we leverage a multidata causal discovery framework with a replicated dataset based on Statistical Hurricane Intensity Prediction Scheme (SHIPS) using ERA5 meteorological reanalysis. We conduct experiments to identify and select predictors causally linked to TC intensity changes. We then train multiple linear regression models to compare causal feature selection with correlation, random forest feature importance, and no feature selection, across five forecast lead times from 1 to 5 days (24 to 120 hours). Causal feature selection consistently outperforms on unseen test cases, especially for lead times shorter than 3 days. Top causal features include vertical shear, mid-tropospheric potential vorticity and surface moisture conditions, which are physically significant yet often underutilized in TC intensity predictions. We build an extended predictor set (SHIPS+) by adding selected features to the standard SHIPS predictors. SHIPS+ yields increased short-term predictive skill at lead times of 24, 48, and 72 hours. Adding nonlinearity using a multilayer perceptron further extends skill to longer lead times, despite our framework being purely regional and not requiring global forecast data. Operational SHIPS tests confirm that three of the six added causally discovered predictors improve forecast skill, with the largest gains at longer lead times. Our results demonstrate that causal discovery improves TC intensity prediction and pave the way toward more empirical forecasts.
Tristan H. Abbott, Timothy W. Cronin, Tom Beucler
Tropical precipitation extremes are expected to strengthen with warming, but quantitative estimates remain uncertain because of a poor understanding of changes in convective dynamics. This uncertainty is addressed here by analyzing idealized convection-permitting simulations of radiative-convective equilibrium in long-channel geometry. Across a wide range of climates, the thermodynamic contribution to changes in instantaneous precipitation extremes follows near-surface moisture, and the dynamic contribution is positive and small, but sensitive to domain size. The shapes of mass flux profiles associated with precipitation extremes are determined by conditional sampling that favors strong vertical motion at levels where the vertical saturation specific humidity gradient is large, and mass flux profiles collapse to a common shape across climates when plotted in a moisture-based vertical coordinate. The collapse, robust to changes in microphysics and turbulence schemes, implies a thermodynamic contribution that scales with near-surface moisture despite substantial convergence aloft and allows the dynamic contribution to be defined by the pressure velocity at a single level. Linking the simplified dynamic mode to vertical velocities from entraining plume models reveals that the small dynamic mode in channel simulations (<~2 %/K) is caused by opposing height-dependences of vertical velocity and density, together with the buffering influence of cloud-base buoyancies that vary little with surface temperature. These results reinforce an emerging picture of the response of extreme tropical precipitation rates to warming: a thermodynamic mode of about 7 %/K dominates, with a minor contribution from changes in dynamics.
Noah D. Brenowitz, Tom Beucler, Michael Pritchard, Christopher S. Bretherton
Neural networks are a promising technique for parameterizing sub-grid-scale physics (e.g. moist atmospheric convection) in coarse-resolution climate models, but their lack of interpretability and reliability prevents widespread adoption. For instance, it is not fully understood why neural network parameterizations often cause dramatic instability when coupled to atmospheric fluid dynamics. This paper introduces tools for interpreting their behavior that are customized to the parameterization task. First, we assess the nonlinear sensitivity of a neural network to lower-tropospheric stability and the mid-tropospheric moisture, two widely-studied controls of moist convection. Second, we couple the linearized response functions of these neural networks to simplified gravity-wave dynamics, and analytically diagnose the corresponding phase speeds, growth rates, wavelengths, and spatial structures. To demonstrate their versatility, these techniques are tested on two sets of neural networks, one trained with a super-parametrized version of the Community Atmosphere Model (SPCAM) and the second with a near-global cloud-resolving model (GCRM). Even though the SPCAM simulation has a warmer climate than the cloud-resolving model, both neural networks predict stronger heating/drying in moist and unstable environments, which is consistent with observations. Moreover, the spectral analysis can predict that instability occurs when GCMs are coupled to networks that support gravity waves that are unstable and have phase speeds larger than 5 m/s. In contrast, standing unstable modes do not cause catastrophic instability. Using these tools, differences between the SPCAM- vs. GCRM- trained neural networks are analyzed, and strategies to incrementally improve both of their coupled online performance unveiled.