An official website of the United States government
Abstract Subseasonal tropical cyclone (TC) reforecasts from the Community Earth System Model version 2 (CAM6) subseasonal prediction system are examined in this study. We evaluate the modeled TC climatology and the probabilistic forecast skill of basin‐wide TC genesis at weekly temporal resolution. Prediction skill is calculated using the Brier skill score relative to a constant annual mean climatology and to a monthly varying seasonal climatology during TC season. The model captures the observed basin‐wide climatological TC seasonality and spatial distributions at weeks 1–6, but TC genesis is largely underestimated from Week 2 onward. For some basins and lead times, the predicted TC genesis is primarily controlled by the number of TC “seeds” and the mean‐state climate condition. The model has good prediction skill relative to the constant climatology across all the basins and lead times, but is only skillful in the eastern Pacific, North Indian Ocean, and Southern Hemisphere at Week 1 when compared to the seasonal climatology, indicating limited skill in predicting deviations from the seasonal cycle. We find strong modulations of the predicted TC genesis at up to 3 weeks of forecast lead time by the Madden‐Julian Oscillation. The interannual variability of predicted TC genesis and accumulated cyclone energy are skillfully predicted in the North Atlantic and the Northwestern Pacific, with a strong modulation by the El Nino‐Southern Oscillation.
more »
« less
Predicting Tropical Cyclone Formation with Deep Learning
Abstract Exploring new techniques to improve the prediction of tropical cyclone (TC) formation is essential for operational practice. Using convolutional neural networks, this study shows that deep learning can provide a promising capability for predicting TC formation from a given set of large-scale environments at certain forecast lead times. Specifically, two common deep-learning architectures including the residual net (ResNet) and UNet are used to examine TC formation in the Pacific Ocean. With a set of large-scale environments extracted from the NCEP–NCAR reanalysis during 2008–21 as input and the TC labels obtained from the best track data, we show that both ResNet and UNet reach their maximum forecast skill at the 12–18-h forecast lead time. Moreover, both architectures perform best when using a large domain covering most of the Pacific Ocean for input data, as compared to a smaller subdomain in the western Pacific. Given its ability to provide additional information about TC formation location, UNet performs generally worse than ResNet across the accuracy metrics. The deep learning approach in this study presents an alternative way to predict TC formation beyond the traditional vortex-tracking methods in the current numerical weather prediction. Significance StatementThis study presents a new approach for predicting tropical cyclone (TC) formation based on deep learning (DL). Using two common DL architectures in visualization research and a set of large-scale environments in the Pacific Ocean extracted from the reanalysis data, we show that DL has an optimal capability of predicting TC formation at the 12–18-h lead time. Examining the DL performance for different domain sizes shows that the use of a large domain size for input data can help capture some far-field information needed for predicting TCG. The DL approach in this study demonstrates an alternative way to predict or detect TC formation beyond the traditional vortex-tracking methods used in the current numerical weather prediction.
more »
« less
- Award ID(s):
- 2309929
- PAR ID:
- 10486198
- Publisher / Repository:
- American Meteorological Society
- Date Published:
- Journal Name:
- Weather and Forecasting
- Volume:
- 39
- Issue:
- 1
- ISSN:
- 0882-8156
- Format(s):
- Medium: X Size: p. 241-258
- Size(s):
- p. 241-258
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
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.more » « less
-
Abstract Applications of machine learning (ML) in atmospheric science have been rapidly growing. To facilitate the development of ML models for tropical cyclone (TC) research, this binary dataset contains a specific customization of the National Center for Environmental Prediction (NCEP)/final analysis (FNL) data, in which key environmental conditions relevant to TC formation are extracted for a range of lead times (0–72 hours) during 1999–2023. The dataset is designed as multi-channel images centered on TC formation locations, with a positive and negative directory structure that can be readily read from any ML applications or common data interface. With its standard structure, this dataset provides users with a unique opportunity to conduct ML application research on TC formation as well as related predictability at different forecast lead times.more » « less
-
Rapid Intensification (RI) in Tropical Cyclone (TC) development is one of the most difficult and still challenging tasks in weather forecasting. In addition to the dynamical numerical simulations, commonly used techniques for RI (as well as TC intensity changes) analysis and prediction are the composite analysis and statistical models based on features derived from the composite analysis. Quite a large number of such selected and pre-determined features related to TC intensity change and RI have been accumulated by the domain scientists, such as those in the widely used SHIPS (Statistical Hurricane Intensity Prediction Scheme) database. Moreover, new features are still being added with new algorithms and/or newly available datasets. However, there are very few unified frameworks for systematically distilling features from a comprehensive data source. One such unified Artificial Intelligence (AI) system was developed for deriving features from TC centers, and here, we expand that system to large-scale environmental condition. In this study, we implemented a deep learning algorithm, the Convolutional Neural Network (CNN), to the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-Interim reanalysis data and identified and refined potentially new features relevant to RI such as specific humidity in east or northeast, vorticity and horizontal wind in north and south relative to the TC centers, as well as ozone at high altitudes that could help the prediction and understanding of the occurrence of RI based on the deep learning network (named TCNET in this study). By combining the newly derived features and the features from the SHIPS database, the RI prediction performance can be improved by 43%, 23%, and 30% in terms of Kappa, probability of detection (POD), and false alarm rate (FAR) against the same modern classification model but with the SHIPS inputs only.more » « less
-
Abstract Heatwaves are projected to increase in frequency and severity with global warming. Improved warning systems would help reduce the associated loss of lives, wildfires, power disruptions, and reduction in crop yields. In this work, we explore the potential for deep learning systems trained on historical data to forecast extreme heat on short, medium and subseasonal time scales. To this purpose, we train a set of neural weather models (NWMs) with convolutional architectures to forecast surface temperature anomalies globally, 1 to 28 days ahead, at ∼200-km resolution and on the cubed sphere. The NWMs are trained using the ERA5 reanalysis product and a set of candidate loss functions, including the mean-square error and exponential losses targeting extremes. We find that training models to minimize custom losses tailored to emphasize extremes leads to significant skill improvements in the heatwave prediction task, relative to NWMs trained on the mean-square-error loss. This improvement is accomplished with almost no skill reduction in the general temperature prediction task, and it can be efficiently realized through transfer learning, by retraining NWMs with the custom losses for a few epochs. In addition, we find that the use of a symmetric exponential loss reduces the smoothing of NWM forecasts with lead time. Our best NWM is able to outperform persistence in a regressive sense for all lead times and temperature anomaly thresholds considered, and shows positive regressive skill relative to the ECMWF subseasonal-to-seasonal control forecast after 2 weeks. Significance StatementHeatwaves are projected to become stronger and more frequent as a result of global warming. Accurate forecasting of these events would enable the implementation of effective mitigation strategies. Here we analyze the forecast accuracy of artificial intelligence systems trained on historical surface temperature data to predict extreme heat events globally, 1 to 28 days ahead. We find that artificial intelligence systems trained to focus on extreme temperatures are significantly more accurate at predicting heatwaves than systems trained to minimize errors in surface temperatures and remain equally skillful at predicting moderate temperatures. Furthermore, the extreme-focused systems compete with state-of-the-art physics-based forecast systems in the subseasonal range, while incurring a much lower computational cost.more » « less
