Search for: All records

Abstract The development of deep learning (DL) weather forecasting models has made rapid progress and achieved comparable or better skill than traditional Numerical Weather prediction (NWP) models, which are generally computationally intensive. However, applications of these DL models have yet to be fully explored, including for severe convective events. We evaluate the DL model Pangu‐Weather in forecasting tornadic environments with one‐day lead times using convective available potential energy (CAPE), 0–6 bulk wind difference (BWD6), and 0–3 km storm‐relative helicity (SRH3). We also compare its performance to the National Centers for Environmental Prediction (NCEP)'s Global Forecast System (GFS), a traditional NWP model. Pangu‐Weather generally outperforms GFS in predicting BWD6 and SRH3 at the closest grid point and hour of the storm report. However, Pangu‐Weather tends to underpredict the maximum values of all convective parameters in the 1–2 hr before the storm across the surrounding grid points compared to the GFS. 

Abstract Improving the skill of medium-range (3–8 day) severe weather prediction is crucial for mitigating societal impacts. This study introduces a novel approach leveraging decoder-only transformer networks to post-process AI-based weather forecasts, specifically from the Pangu-Weather model, for improved severe weather guidance. Unlike traditional post-processing methods that use a dense neural network to predict the probability of severe weather using discrete forecast samples, our method treats forecast lead times as sequential “tokens”, enabling the transformer to learn complex temporal relationships within the evolving atmospheric state. We compare this approach against post-processing of the Global Forecast System (GFS) using both a traditional dense neural network and our transformer, as well as configurations that exclude convective parameters to fairly evaluate the impact of using the Pangu-Weather AI model. Results demonstrate that the transformer-based post-processing significantly enhances forecast skill compared to dense neural networks. Furthermore, AI-driven forecasts, particularly Pangu-Weather initialized from high resolution analysis, exhibit superior performance to GFS in the medium-range, even without explicit convective parameters. Our approach offers improved accuracy, and reliability, which also provides interpretability through feature attribution analysis, advancing medium-range severe weather prediction capabilities. 

Abstract The National Weather Service plays a critical role in alerting the public when dangerous weather occurs. Tornado warnings are one of the most publicly visible products the NWS issues given the large societal impacts tornadoes can have. Understanding the performance of these warnings is crucial for providing adequate warning during tornadic events and improving overall warning performance. This study aims to understand warning performance during the lifetimes of individual storms (specifically in terms of probability of detection and lead time). For example, does probability of detection vary based on if the tornado was the first produced by the storm, or the last? We use tornado outbreak data from 2008 to 2014, archived NEXRAD radar data, and the NWS verification database to associate each tornado report with a storm object. This approach allows for an analysis of warning performance based on the chronological order of tornado occurrence within each storm. Results show that the probability of detection and lead time increase with later tornadoes in the storm; the first tornadoes of each storm are less likely to be warned and on average have less lead time. Probability of detection also decreases overnight, especially for first tornadoes and storms that only produce one tornado. These results are important for understanding how tornado warning performance varies during individual storm life cycles and how upstream forecast products (e.g., Storm Prediction Center tornado watches, mesoscale discussions, etc.) may increase warning confidence for the first tornado produced by each storm. Significance StatementIn this study, we focus on better understanding real-time tornado warning performance on a storm-by-storm basis. This approach allows us to examine how warning performance can change based on the order of each tornado within its parent storm. Using tornado reports, warning products, and radar data during tornado outbreaks from 2008 to 2014, we find that probability of detection and lead time increase with later tornadoes produced by the same storm. In other words, for storms that produce multiple tornadoes, thefirsttornado is generally the least likely to be warned in advance; when it is warned in advance, it generally contains less lead time than subsequent tornadoes. These findings provide important new analyses of tornado warning performance, particularly for the first tornado of each storm, and will help inform strategies for improving warning performance. 

Abstract This study explores how tornadic supercell soundings significantly differ from the same‐location and same‐hour baseline environment soundings, sampled from the days prior to or following the event. Permutation testing is used to identify whether sounding‐derived parameters mixed‐layer convective available potential energy and 0–1 km storm‐relative helicity are significantly different between the tornadic and baseline environment. Typically, in an environment with marginal values of certain key environmental parameters, anomalous values of those environmental parameters are more strongly associated with supercell tornadoes. Furthermore, many tornadic events already exhibit environmental conditions favorable for tornadic supercells a day prior to the event itself. Generally, supercell tornadoes that occur during typical peak tornadic activity time frames are easier to distinguish from baseline (non‐tornadic) environments compared to those occurring in other time frames. Spatiotemporal variations of distinguishability between tornadic and baseline environmental parameters add complexity to traditional parameter‐based fixed threshold forecasting.