• Energy Research

The moisture origins and associated physical mechanisms for tornadoes of various climatic regions of the United States were investigated. The NOAA Air Resources Laboratory HYSPLIT model and a moisture attribution algorithm were used in conjunction with statistical analyses to explore these relationships on a climate scale (1981–2017). It was found that moisture sources of United States tornadic convection exhibit distinctive regionality. For example, moisture contributions to east coast tornadoes primarily emanate from the Atlantic Ocean as opposed to the Gulf of Mexico. Moisture sources of a class of non-significant severe thunderstorms also show regionality based on the location of storm occurrence with latitudinal differences to that of tornadic storms. Moisture increases for tornadic and non-significant severe storms were shown to be closely related to the respective air parcel's temperature, with underlying sea surface temperatures playing a less important role. Long-term trends of moisture uptake and horizontal advection were also explored using the Mann-Kendall test and linear regressions. Both statistical analyses demonstrate that the magnitude of advective fluxes and the rate of moisture increases have been rising since the 1980s, which can increase the complexity of forecasting downstream convection. The potential ramifications of these trends on storm development and prediction are discussed to conclude the article.

AbstractThis is a test case study assessing the ability of deep learning methods to generalize to a future climate (end of 21st century) when trained to classify thunderstorms in model output representative of the present‐day climate. A convolutional neural network (CNN) was trained to classify strongly rotating thunderstorms from a current climate created using the Weather Research and Forecasting model at high‐resolution, then evaluated against thunderstorms from a future climate and found to perform with skill and comparatively in both climates. Despite training with labels derived from a threshold value of a severe thunderstorm diagnostic (updraft helicity), which was not used as an input attribute, the CNN learned physical characteristics of organized convection and environments that are not captured by the diagnostic heuristic. Physical features were not prescribed but rather learned from the data, such as the importance of dry air at mid‐levels for intense thunderstorm development when low‐level moisture is present (i.e., convective available potential energy). Explanation techniques also revealed that thunderstorms classified as strongly rotating are associated with learned rotation signatures. Results show that the creation of synthetic data with ground truth is a viable alternative to human‐labeled data and that a CNN is able to generalize a target using learned features that would be difficult to encode due to spatial complexity. Most importantly, results from this study show that deep learning is capable of generalizing to future climate extremes and can exhibit out‐of‐sample robustness with hyperparameter tuning in certain applications.