An official website of the United States government
Abstract A primary goal of the National Oceanic and Atmospheric Administration Warn-on-Forecast (WoF) project is to provide rapidly updating probabilistic guidance to human forecasters for short-term (e.g., 0–3 h) severe weather forecasts. Postprocessing is required to maximize the usefulness of probabilistic guidance from an ensemble of convection-allowing model forecasts. Machine learning (ML) models have become popular methods for postprocessing severe weather guidance since they can leverage numerous variables to discover useful patterns in complex datasets. In this study, we develop and evaluate a series of ML models to produce calibrated, probabilistic severe weather guidance from WoF System (WoFS) output. Our dataset includes WoFS ensemble forecasts available every 5 min out to 150 min of lead time from the 2017–19 NOAA Hazardous Weather Testbed Spring Forecasting Experiments (81 dates). Using a novel ensemble storm-track identification method, we extracted three sets of predictors from the WoFS forecasts: intrastorm state variables, near-storm environment variables, and morphological attributes of the ensemble storm tracks. We then trained random forests, gradient-boosted trees, and logistic regression algorithms to predict which WoFS 30-min ensemble storm tracks will overlap a tornado, severe hail, and/or severe wind report. To provide rigorous baselines against which to evaluate the skill of the ML models, we extracted the ensemble probabilities of hazard-relevant WoFS variables exceeding tuned thresholds from each ensemble storm track. The three ML algorithms discriminated well for all three hazards and produced more reliable probabilities than the baseline predictions. Overall, the results suggest that ML-based postprocessing of dynamical ensemble output can improve short-term, storm-scale severe weather probabilistic guidance.
more »
« less
Gridded Severe Hail Nowcasting Using 3D U-Nets, Lightning Observations, and the Warn-on-Forecast System
Abstract Hailstorms cause billions of dollars in damage across the United States each year. Part of this cost could be reduced by increasing warning lead times. To contribute to this effort, we developed a nowcasting machine learning model that uses a 3D U-Net to produce gridded severe hail nowcasts for up to 40 min in advance. The three U-Net dimensions uniquely incorporate one temporal and two spatial dimensions. Our predictors consist of a combination of output from the National Severe Storms Laboratory Warn-on-Forecast System (WoFS) numerical weather prediction ensemble and remote sensing observations from Vaisala’s National Lightning Detection Network (NLDN). Ground truth for prediction was derived from the maximum expected size of hail calculated from the gridded NEXRAD WSR-88D radar (GridRad) dataset. Our U-Net was evaluated by comparing its test set performance against rigorous hail nowcasting baselines. These baselines included WoFS ensemble Hail and Cloud Growth Model (HAILCAST) and a logistic regression model trained on WoFS 2–5-km updraft helicity. The 3D U-Net outperformed both these baselines for all forecast period time steps. Its predictions yielded a neighborhood maximum critical success index (max CSI) of ∼0.48 and ∼0.30 at forecast minutes 20 and 40, respectively. These max CSIs exceeded the ensemble HAILCAST max CSIs by as much as ∼0.35. The NLDN observations were found to increase the U-Net performance by more than a factor of 4 at some time steps. This system has shown success when nowcasting hail during complex severe weather events, and if used in an operational environment, may prove valuable.
more »
« less
- Award ID(s):
- 2019758
- PAR ID:
- 10567244
- Publisher / Repository:
- AMS
- Date Published:
- Journal Name:
- Artificial Intelligence for the Earth Systems
- Volume:
- 3
- Issue:
- 4
- ISSN:
- 2769-7525
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract The quantification of storm updrafts remains unavailable for operational forecasting despite their inherent importance to convection and its associated severe weather hazards. Updraft proxies, like overshooting top area from satellite images, have been linked to severe weather hazards but only relate to a limited portion of the total storm updraft. This study investigates if a machine learning model, namely, U-Nets, can skillfully retrieve maximum vertical velocity and its areal extent from three-dimensional gridded radar reflectivity alone. The machine learning model is trained using simulated radar reflectivity and vertical velocity from the National Severe Storm Laboratory’s convection permitting Warn-on-Forecast System (WoFS). A parametric regression technique using the sinh–arcsinh–normal distribution is adapted to run with U-Nets, allowing for both deterministic and probabilistic predictions of maximum vertical velocity. The best models after hyperparameter search provided less than 50% root mean squared error, a coefficient of determination greater than 0.65, and an intersection over union (IoU) of more than 0.45 on the independent test set composed of WoFS data. Beyond the WoFS analysis, a case study was conducted using real radar data and corresponding dual-Doppler analyses of vertical velocity within a supercell. The U-Net consistently underestimates the dual-Doppler updraft speed estimates by 50%. Meanwhile, the area of the 5 and 10 m s−1updraft cores shows an IoU of 0.25. While the above statistics are not exceptional, the machine learning model enables quick distillation of 3D radar data that is related to the maximum vertical velocity, which could be useful in assessing a storm’s severe potential. Significance StatementAll convective storm hazards (tornadoes, hail, heavy rain, straight line winds) can be related to a storm’s updraft. Yet, there is no direct measurement of updraft speed or area available for forecasters to make their warning decisions from. This paper addresses the lack of observational data by providing a machine learning solution that skillfully estimates the maximum updraft speed within storms from only the radar reflectivity 3D structure. After further vetting the machine learning solutions on additional real-world examples, the estimated storm updrafts will hopefully provide forecasters with an added tool to help diagnose a storm’s hazard potential more accurately.more » « less
-
Abstract This study compares real-time forecasts produced by the Warn-on-Forecast System (WoFS) and a hybrid ensemble and variational data assimilation and prediction system (WoF-Hybrid) for 31 events during 2021. Object-based verification is used to quantify and compare strengths and weaknesses of WoFS ensemble forecasts with 3-km horizontal grid spacing and WoF-Hybrid deterministic forecasts with 1.5-km horizontal grid spacing. The goal of such comparison is to provide evidence as to whether WoF-Hybrid has performance characteristics that complement or improve upon those of WoFS. Results indicate that both systems provide similar accuracy for timing and placement of thunderstorm objects identified using simulated reflectivity. WoF-Hybrid provides more accurate forecasts of updraft helicity tracks. Differences in forecast quality are case dependent; the largest difference in accuracy favoring WoF-Hybrid occurs in eight cases identified as “high-impact” by the quantity of National Weather Service Local Storm Reports, while WoFS performance is favored at short lead times for 10 “moderate-” and 13 “low-impact” events. WoF-Hybrid reflectivity objects are closer in size and location to observed objects. However, a higher thunderstorm overprediction bias is identified in WoF-Hybrid, particularly early in the forecast. Two severe weather events are selected for detailed investigation. In the case of 26 May, both systems had similar skill; however, for 10 December, WoF-Hybrid forecasts significantly outperformed WoFS forecasts. These results show improved performance for WoF-Hybrid over WoFS under certain regimes that warrants further investigation. To understand reasons for these differences will help incorporate higher-resolution modeling into Warn-on-Forecast systems. Significance StatementThe NOAA Warn-on-Forecast (WoF) project uses advanced data assimilation for rapidly updating numerical weather prediction systems to provide forecasts of individual thunderstorms. Forecasts show promise for enabling greater warning lead time for some storms. The flagship Warn-on-Forecast System (WoFS) is a 36-member analysis and 18-member forecast system at 3-km grid spacing. The project also produced a single member system that employs variational analysis and produces a deterministic forecast at 1.5-km grid spacing (WoF-Hybrid). This study seeks to evaluate and compare the performance of WoFS and WoF-Hybrid for 31 severe weather events that occurred during 2021. Results found that WoF-Hybrid predicts storm rotation particularly well compared to WoFS, and several other strengths and limitations of both systems are identified. This research may help us understand the complementary nature of two systems and improve our ability to provide more reliable 0–6-h forecasts in the future.more » « less
-
null (Ed.)Abstract As lightning-detection records lengthen and the efficiency of severe weather reporting increases, more accurate climatologies of convective hazards can be constructed. In this study we aggregate flashes from the National Lightning Detection Network (NLDN) and Arrival Time Difference long-range lightning detection network (ATDnet) with severe weather reports from the European Severe Weather Database (ESWD) and Storm Prediction Center (SPC) Storm Data on a common grid of 0.25° and 1-h steps. Each year approximately 75–200 thunderstorm hours occur over the southwestern, central, and eastern United States, with a peak over Florida (200–250 h). The activity over the majority of Europe ranges from 15 to 100 h, with peaks over Italy and mountains (Pyrenees, Alps, Carpathians, Dinaric Alps; 100–150 h). The highest convective activity over continental Europe occurs during summer and over the Mediterranean during autumn. The United States peak for tornadoes and large hail reports is in spring, preceding the maximum of lightning and severe wind reports by 1–2 months. Convective hazards occur typically in the late afternoon, with the exception of the Midwest and Great Plains, where mesoscale convective systems shift the peak lightning threat to the night. The severe wind threat is delayed by 1–2 h compared to hail and tornadoes. The fraction of nocturnal lightning over land ranges from 15% to 30% with the lowest values observed over Florida and mountains (~10%). Wintertime lightning shares the highest fraction of severe weather. Compared to Europe, extreme events are considerably more frequent over the United States, with maximum activity over the Great Plains. However, the threat over Europe should not be underestimated, as severe weather outbreaks with damaging winds, very large hail, and significant tornadoes occasionally occur over densely populated areas.more » « less
-
Abstract An ensemble postprocessing method is developed for the probabilistic prediction of severe weather (tornadoes, hail, and wind gusts) over the conterminous United States (CONUS). The method combines conditional generative adversarial networks (CGANs), a type of deep generative model, with a convolutional neural network (CNN) to postprocess convection-allowing model (CAM) forecasts. The CGANs are designed to create synthetic ensemble members from deterministic CAM forecasts, and their outputs are processed by the CNN to estimate the probability of severe weather. The method is tested using High-Resolution Rapid Refresh (HRRR) 1–24-h forecasts as inputs and Storm Prediction Center (SPC) severe weather reports as targets. The method produced skillful predictions with up to 20% Brier skill score (BSS) increases compared to other neural-network-based reference methods using a testing dataset of HRRR forecasts in 2021. For the evaluation of uncertainty quantification, the method is overconfident but produces meaningful ensemble spreads that can distinguish good and bad forecasts. The quality of CGAN outputs is also evaluated. Results show that the CGAN outputs behave similarly to a numerical ensemble; they preserved the intervariable correlations and the contribution of influential predictors as in the original HRRR forecasts. This work provides a novel approach to postprocess CAM output using neural networks that can be applied to severe weather prediction. Significance StatementWe use a new machine learning (ML) technique to generate probabilistic forecasts of convective weather hazards, such as tornadoes and hailstorms, with the output from high-resolution numerical weather model forecasts. The new ML system generates an ensemble of synthetic forecast fields from a single forecast, which are then used to train ML models for convective hazard prediction. Using this ML-generated ensemble for training leads to improvements of 10%–20% in severe weather forecast skills compared to using other ML algorithms that use only output from the single forecast. This work is unique in that it explores the use of ML methods for producing synthetic forecasts of convective storm events and using these to train ML systems for high-impact convective weather prediction.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1175/AIES-D-24-0026.1
